Versions & what's new

Mandatory update — please install promptly; this update is required. Minimum version — older builds are no longer supported.

2026.08.19.191919 Aug 2026Current

• FixedTables say again what is in their columns. In the ticket list, 3 of 13 columns carried a heading; across eleven lists there sat 35 columns sat beneath an empty header bar. The heading was not missing — it was configured, translated and complete; it simply never reached the column. Anyone sorting a list or mapping a column had to work out from the values what they meant. The row-actions column now carries a heading too: a header cell without text stays silent for screen readers.
• NewSeventeen more lists now have a header row. Not every table in the client is a grid — a whole range of lists draws its own table layout, and there the heading was not merely missing, it never existed at all. Affected were 17 places across 12 pages: contract, ticket and lifecycle references on the device, the field list of the class editor, open tickets and the impact analysis on the configuration item, the assignments in the GVP, the messenger status overview, two satisfaction distributions, the key figures of the process documentation, the severity distribution of the vulnerabilities, the most frequent prefixes in the IPAM overview, the permission rules and „My applications“.
• ImprovedNo more duplicate labels in the permission rules. Beneath every value, small grey type repeated which field it was. As long as the list had no header row, that was its only label and therefore necessary; now it stands once at the top, and the repetition in every row is gone.

2026.08.19.145219 Aug 2026

• NewThe interface now keeps itself up to date in all areas. What began with the service desk now applies to the whole application: configuration items, assets, licences, contracts, network addresses and discovery — security, ISMS, data protection, business continuity and backup — organisation, processes, task boards and approvals — users, roles and reports. 159 views across client and web portal, up from 38. This counts for most where the change does not come from a person at all: a discovery run, a CSV import or a nightly backup job writes while someone has the list open — and the list shows it without them doing anything. What is still reported is only that something has changed, never the content; the page reloads by itself, and the access check applies as always.
• NewLeaving a form with unsaved input now asks first. Until now not a single one of the 144 client views did that: one click in the menu and half an hour of typing was gone, without a warning and without a way back. Every page and every dialog with a save action now asks — in client and web portal with the same wording. Pages with several independent sections track the state per section: saving one no longer marks the unsaved input in another as done.
• ImprovedA live notice no longer overwrites input. The two together — a page that updates itself and an open form — would otherwise have been the most expensive case of all: a colleague’s first notice would have replaced the text while someone was typing. Detail pages therefore hold back while editing is in progress, lists while rows are selected, and the process editor shows a notice instead of silently swapping the canvas.
• ImprovedA selection survives a reload. Seven pages always picked the first entry when opening — for a stocktake, a network range or a continuity scope that is right. On a reload it would have been the mistake: the page would have jumped away under someone’s hands as soon as a new run started somewhere.
• ImprovedA running import no longer wakes open lists row by row. An import writing row by row would have turned one operation into hundreds of notices. The server now groups them over a short time window — measured: 500 write operations produce 2 notices instead of 500. The last state is always reported; a list missing its final row would be worse than one that never reloads — it would look current.
• NewTwelve lists in the client now share the same table. Column choice, sorting, grouping, search and export on every one of them, instead of something different on each. Converting them showed why this is more than cosmetics: in twenty places the column header and the export field had drifted apart — six in the ticket list alone, where the affected columns were therefore also missing from the export.
• ImprovedEvery dialog can be operated from the keyboard. In 106 notice windows the Enter key did nothing — you had to reach for the mouse to confirm “Understood”. One dialog could not be closed with Esc at all, because “Cancel” sat on the wrong button. Both fixed, and all dialogs now follow one shared building block instead of 403 individual ones.
• ImprovedStatus information looks the same everywhere. In the web portal five places coloured the same status differently, and in two places the raw server value was on screen — text that no translation would ever have reached. There is now one mapping for all thirteen areas.
• ImprovedBulk actions in the web portal. Where the server offers bulk operations, rows can now be selected and processed together — and only there. A checkbox without effect is worse than none.
• FixedThe settings page’s collective save button wrote more than it should. It sat in the page header, visibly belonged to nothing and on every click saved all areas — the display setting even twice, and the AD/LDAP configuration was rewritten unasked, even while you were working on the SLA settings. Every area now has its own save button on its own card.

2026.08.18.204318 Aug 2026

• NewChanges made by others become visible without reloading the page. Until now, an open page only refreshed when you yourself changed something. Whatever a colleague, an import or a background run wrote was only visible after switching pages — and because navigating away and back worked, it felt like a quirk rather than a gap. The server now reports every saved change from one place, regardless of what triggered it: the interface, an import, incoming mail or a scheduled run. The ticket list, the ticket detail view and the service desk workspace are connected to it in both the client and the web portal; the remaining areas will follow. Only the fact that something changed is reported — never the content. The page reloads by itself, and the access check applies as always.
• FixedThe live push in the desktop client fell silent after fifteen minutes — permanently. The access token of the live connection was collected at sign-in once . It lives for 15 minutes; after the first reconnect — a network drop, standby, a change of Wi-Fi — the connection was dead for good until the application was restarted. The status bar kept showing „reconnecting …“ forever. The token is now fetched fresh on every connection attempt, and switching tenants renews the connection.
• FixedThe start page never received a notification. The personal dashboard was meant to refresh itself on every notification — the registration required for that was never called anywhere in the code base. It lived on its 60-second timer alone. Also removed: three receivers for events no server has ever sent. They looked finished in the code and did nothing.
• FixedSwitching a module on or off only took effect after a restart. Anyone toggling a module in the administration still saw the menu entry — or still did not. In the web portal only a full page reload helped, in the client only restarting the application. Both now take effect immediately.
• ImprovedNo more deleting without a confirmation. 35 paths across the client and the web portal removed data on a single click, irrecoverably. They now ask first — on both interfaces with the same wording and with the default on „Cancel“, so that the Enter key does not confirm the very thing being warned about.
• ImprovedDelete buttons are recognisable as such — and legible. Four out of five delete buttons in the client carried no warning colour. The one warning style that did exist was itself coloured wrongly: its text colour belonged to a different colour role and missed the contrast requirement — of all things, the button that warns about data loss was the hardest one to read.
• FixedTwelve column headers accepted the click and did not sort. A sort header without effect is worse than a table without sorting — the user looks for the mistake in themselves. Where sorting is missing, it is no longer offered.
• ImprovedA loading error now says that it is one. When a request failed, the table said literally „No data available“ — a statement about the data, where one about the connection or the permission was meant. The standard table now tells the two apart and offers „Try again“; the service desk workspace already uses it.

2026.08.18.131518 Aug 2026

• SecurityApproval rules of other tenants were read — and overwritten. The settings for CAB approval of a change request were fetched without any condition from the table: in shared-database operation therefore the first row, no matter which organisation it belonged to. It is precisely this value that decides who may approve and whether self-approval is permitted — and the save path wrote back into that same foreign row. One organisation could thereby lift another's four-eyes principle. Fixed, additionally secured by a tenant filter and proven by dedicated tests. This update is therefore mandatory.
• SecurityThe portal showed the knowledge articles of all tenants. The public knowledge base is deliberately reachable without signing in — and that is exactly where the tenant filter does not bite, because there is no signed-in session. The only reference was a voluntary parameter that the portal never sent. The tenant is now determined on the server; if it is not unambiguous, nothing is delivered rather than too much. The related articles carry the boundary as well — otherwise the entry point would have been filtered and the second click no longer.
• NewApplications and the people responsible for them. Not every piece of software is an application: the software inventory shows what is installed while the new application overview shows what the organisation manages as a business application — with product owner, deputy, department, support and the servers behind it. Each application has at most one product owner and at most one deputy; otherwise nobody decides when it counts. The tab „Not managed“ counts what is not yet assigned to any application — that is the responsibility of IT, made visible instead of kept as an unwritten rule. And „My applications“ shows every account what it has been entered for: a responsibility the responsible person knows nothing about is none.
• FixedService requests could no longer be resolved. Resolving is blocked while fulfilment tasks are open — but no interface showed these tasks, and the shipped sample data creates four of them for every open service request. The message therefore named tasks that were nowhere to be seen and nowhere to be ticked off. The ticket now carries a task card: create, assign, tick off, sort.
• NewFields for public administration can be maintained. File reference, statutory deadline, origin channel and the link to the e-file had long been kept by the server, and the statutory deadline monitored separately from the SLA — only no workplace knew these fields. For public authorities the most visible part of the module, until now reachable only through the raw interface.
• NewSupplier deadline on the ticket. Supplier, contract reference and the promised deadline can now be set. The monitoring for it already existed — it could simply never fire, because there was no way to set the deadline.
• NewCalls visible on the ticket. Incoming and outgoing calls with number, duration and note were recorded, but shown by no interface.
• FixedThe escalation ran into the void. The support group selection on the ticket was a hard-wired list of twelve names — the groups actually maintained were never loaded, and only the name was stored, never the identifier. The deadline monitoring resolves the escalation recipient by name and therefore never found one. The selection now comes from the actual data; support groups are also shipped during setup — the list was always empty at first.
• ImprovedThe status selection now offers only what is possible. It always showed all seven states, regardless of the current one: from „In progress“ it offered „New“ and „Closed“, both of which the server rejects. The error was only visible after the click, as a message. The selection now asks the server for the permitted transitions instead of rebuilding them a second time — two tables of the same matter drift apart.
• ImprovedThe CMDB menu has been recut. It carried fifteen entries and thus four different activities side by side: keeping inventory, bringing data in, maintaining master data, software. Now seven entries, next to them the groups „Applications & software“, „Capture & reconciliation“ and „Catalogues & templates“. No entry is dropped, none is renamed — they simply sit where people look for them. The portal follows the same layout, as far as it has the pages.
• ImprovedThe service desk speaks the configured language. Eighty-one labels stood fixed in German in the program — the status filter of the ticket list, the buckets of the data quality view, the support group dialogues and the prompts in the ticket. In every other language they stayed German, without error and without warning.
• ImprovedManual: chapter „Product security & reporting channels“ can be found. The chapter was built and shipped, but was not listed in the document map and not named in the English and French cross-references. Also new: the section on applications and responsibilities in DE, EN and FR.
• FixedEvery status change pulled an empty table along. The original SLA timer had been without function since the switch to markers on the ticket — nobody ever wrote into its table, yet every status change loaded it. It is dropped. The migration aborts instead of deleting, should an installation unexpectedly still hold rows.

2026.08.18.083918 Aug 2026

• SecurityCI classes get their tenant separation. Out of 56 CMDB entities, the CI class was the only without a tenant filter — and neither of the two places that create one stamped a tenant on it. Every class an organisation created itself was therefore visible to all others, its full attribute schema retrievable via its identifier, and someone else's class name blocked your own creation with an error message nobody could make sense of. Fixed, and covered by a dedicated test. This update is therefore mandatory.
• NewRack types can be maintained. Form factor, width, rack units, counting direction, mounting depth, load capacity and outer dimensions can now be kept as a catalogue — create, edit, remove. The data had been there for a while, but could only be operated through the raw interface, even though the rack plan's occupancy calculation builds directly on it. In the rack plan itself, „Assign rack type“ transfers the dimensions to a cabinet; without the „Overwrite existing“ tick only empty attributes are filled — assigning a model afterwards does not lose the values measured at the cabinet.
• NewComponent templates on the device model. The equipment a device is meant to inherit when it is created can now be maintained, not merely viewed. The name range sits right next to the name: „Port“ plus „1-48“ produces forty-eight rows — without it, maintaining a switch's port layout by hand would be so tedious that it simply would not happen. Shipped models stay locked, because a catalogue update would overwrite them.
• NewData migration from NetBox is operable. Store the connection and its access token, test the connection, define the class mapping, start the run, complete it — four steps on one page. The dry run is the default and not merely an option: it is the only opportunity to check the mapping table before thousands of objects are created with the wrong class. Objects that cannot be mapped are skipped and listed in the report — never guessed, because a wrong type costs more in a CMDB than a missing one. A data migration is an organisation's very first step; until now it ran only through the raw interface.
• NewSoftware inventory as a view of its own. The software list gathered by discovery — grouped by product, publisher and version, with the affected machines to expand and a jump to the device. It is the basis of both licence reconciliation and vulnerability matching, and until now could only be viewed through the interface.
• ImprovedCommercial data can finally be entered. Purchase price, purchase date, cost centre, warranty start and support contract had, across the entire interface, not a single input field — the interface accepted them, the form was missing. Depreciation therefore always calculated with zero, the cost-centre roll-up stayed empty, and the data-quality view inevitably reported „100 % without a cost centre“ — not because nobody maintained it, but because there was no way to. The fields now sit in the creation wizard and in the detail editor.
• ImprovedLicence master data stays editable. Quantity, price per unit, purchase date, support/maintenance end, contract link and contract reference could not be corrected after creation — on saving they were merely passed through unchanged, which looked like maintenance and was a round trip. Quantity in particular carries the whole over- and under-licensing calculation; the maintenance end feeds the daily early warning, which with an empty field never found anything and looked healthy doing so.
• ImprovedThe journal field in a ticket grows with the text. The input field had a fixed height — with any longer explanation you saw three lines of your own text and scrolled inside the field. It now grows as you type, and the grip below it lets you drag it larger still. A small legend also shows which Markdown formatting the entry understands — it was already rendered that way, nobody just said so.
• ImprovedThe asset, label and licence interfaces are translatable. Lifecycle phases, disposal methods, governance categories, the explanations of the printer languages and the licence governance recommendations were hard-wired in German. In every other language they stayed German — no error, no warning. Licence governance no longer delivers its recommendation as a finished sentence from the server, but as an identifier with figures that the workstation phrases in its own language.
• ImprovedThe counting basis is shown on the licence. What consumption is measured against — per user, per core, per processor, per subscription seat — was stored but shown nowhere; the governance view held a figure whose reference the reader had to guess. It now sits beneath the licence type.
• FixedTwo purchases of the same title were counted twice. Licence reconciliation measured consumption per licence line instead of per product. Buying the same title twice charged the full consumption to both lines: both counted as under-licensed, the purchase recommendation appeared twice and so did its cost — precisely the figure a buying decision rests on. Consumption is now measured once per product and distributed (oldest line first, the surplus onto the newest).
• FixedA price of 12.50 was stored as 1250. Monetary amounts from input fields were parsed against the invariant culture; the German notation with a comma therefore arrived a hundred times too high — without any error message, and from there into true-up costs and depreciation. The amount is now recognised from the position of the characters rather than from a culture setting, covered by dedicated tests.
• FixedThe licence type was discarded on saving. The interface accepted it and silently dropped it, and in the workstation it was bound to no control at all. A type chosen wrongly at creation was therefore permanent — and the creation dialog was fixed on „Subscription“.
• FixedOpen-source software counted as under-licensed. The „not subject to counting“ exception was stored but wired up nowhere. Free software recorded with a quantity of 1 and installed on fifty machines therefore appeared as forty-nine times under-licensed — complete with audit risk and a purchase recommendation.
• FixedThe process map saw no sub-processes. It compared against the technical identifier of the relation, whereas the application stores the label. Every sub-process therefore stood as a root of its own, was not counted towards its parent process and was not indented when expanded. Nothing about it was visible — it looked like a flat process landscape, not like a fault.
• FixedFour places showed a key name instead of a text. In the ABAC rules, the process organisation, the report builder and the weather view, a string such as „Lf913ca67fb“ appeared on screen in one place each. Fixed — and covered by a test that from now on checks every text key in use against the language file.
• FixedThe printer languages of the label media were hard-wired. When a language was added on the server, the workstation did not notice and the selection stayed incomplete. It is now fetched from the server.

2026.08.17.203217 Aug 2026

• NewAlerting — the module goes into service. An alert is now a record of its own, with a recipient group, a delivery log and acknowledgements. What is logged is every attempt and not merely the final outcome: from a single row per recipient with one „result“ field, delivery rates and response times could not be calculated — and those are exactly what the check stage of ISO 22301 requires. Acknowledgements are never overwritten — what counts is the first response. Title and text are held on the alert itself, so that a target object renamed later cannot falsify the log retrospectively. An alert cannot be unsubscribed from: the notification preferences are deliberately not consulted — an alert that can be switched off is not an alert, and the failure would be invisible, because the system would report success while nobody had been reached. A channel that has not been set up is logged as an attempt marked „not configured“ instead of quietly going missing.
• NewVulnerability management gets its user interface. The analysis was already running in the background; until now there was no way to work with it. There are now four views: an overview by severity, by product and by affected device; the findings with assessment (confirmed, false positive, not applicable, accepted — the last three each with a justification, „accepted“ additionally with a deadline, because without both it would not be a decision in an audit); the product mapping with its remaining list and rule editing including a preview before saving, so that an over-broad pattern is noticed beforehand and not only at the next reconciliation; and the feeds with the catalogue status. The view grouped by product is the real lever here: one row „Java 8u201 on 43 machines“ leads to one change request, not to 43. Severity never carries the colour alone — symbol, text and the CVSS value always stand beside it.
• NewBackup: media, off-site storage and retention are now operable. Tapes and removable media can be created and maintained, their readability recorded (explicitly no a substitute for the restore test — the dialogue says so as well) and above all: off-site moves and returns are signed off, with location, time and person. Only then does the „1 off-site“ of the 3-2-1 rule become a fact instead of an intention. Newly configurable is the retention of run data — these two periods determine when backup records disappear irretrievably; until now they existed only in the interface.
• NewMy backup reviews in the Android app. Your own open review periods — including those for which you are entered as deputy — can be viewed and completed on the phone. The app does not recalculate the completion condition itself; it asks the server: a button that is green on the phone and locked at the desk would be worse than none. If it is locked, the reason stands below it.
• ImprovedBackup and information security are fully translatable. Both modules carried hard-wired German text in a number of places — metric labels, feedback messages, calculated figures. In an English or French interface they remained German. That has been cleaned up; the language switch now covers both modules completely.
• ImprovedCMDB bulk editing selects attributes from a list. The attribute name previously had to be typed as free text — a typing error meant the edit matched nothing, and it did not say so either. A selection list is now available.
• FixedA fetch run could inherit the position of another. The marker recording how far a vulnerability feed had already been fetched was held on the service rather than on the individual run. Because all enabled feeds run one after another through the same service, the position of one could live on into the next — and a wrong position skips advisories without anything failing. The position now belongs to the run.
• FixedAndroid app: delivery possible again, notifications arrive. The release build of the app was blocked by a package version conflict in the user-interface library; behind it lay twelve concealed errors, because the vendor had declared the entire device-registration interface obsolete without a replacement. Both have been cleared up. Furthermore: the notification channels are now created at startup instead of at the first alert, tapping a notification leads to the right place even from the closed state, and alerts run through a channel of their own with their own sound and alarm volume — even when ordinary messages are set to silent.

2026.08.17.073217 Aug 2026

• FixedThe Windows client no longer started on machines in a Windows domain. After the splash screen nothing happened; the process could only be ended via Task Manager. It was neither a network nor a server problem. The splash screen deliberately stands for five seconds so that version, server and sign-in account remain readable — and the appearance of the main window was inadvertently tied to that purely visual delay: it only came once the delay had expired. If the delay did not expire, the window did not come, without an error and without a message. This only became noticeable with one setting switched on — sign-in with the Windows account at startup; on machines with a local Windows account — and therefore on every test setup — it never occurred. Exactly why this coupling exists has not been conclusively clarified and is being pursued further. What has been fixed is the structural fault behind it, and that weighs more than the question of what triggered it: the main window now appears immediately and no longer depends on any delay; the five seconds now apply to the splash screen alone. Until you update, there is a workaround: switch off Windows sign-in for that workstation.
• ImprovedThere is now always a way out of the sign-in at startup. The way out of the startup screen only appeared when several connections were configured — that is, precisely not in the most common situation, the single installation. And it was exactly there that a hanging start left nothing but Task Manager. The link is now always present; with a single connection it leads, as Cancel , to the familiar sign-in screen, otherwise as before to the connection picker.
• FixedA failed start now reports itself instead of standing there silently. If an error occurred before the first window, it was logged and then swallowed: the process kept running even though there was nothing left to operate. Now a message names the error and the path of the log file, and the client shuts down in an orderly way. After startup, the previous behaviour explicitly remains — an error on a single page must not end the running session.
• FixedThe database reference in the manual was only up to date in German. The chapter is generated from the running database; until now only the German version was regenerated, while the English and French ones stayed on an older state. All three versions match again.
• NewThe foundation for push notifications on mobile devices is in place. The platform now brings the server side with it: device registration, a queue with retry and a give-up limit, and delivery via a central broker. Routing through a broker is a security decision: otherwise every customer server would hold a service account key that is technically valid for all installations. Let us state the obvious: The mobile app is not yet generally available, and without it this point changes nothing for you today. It is listed because it is part of this version.

Week 33 · 10–16 August 202620 versions

2026.08.16.163716.08.2026

• NewThe software bill of materials is now inside the application. Under System status › Software bill of materials (SBOM) shows what your installation consists of: the Ordivis components, the runtimes and the database with their end of support, all 284 third-party libraries with version, licence and purpose, and the embedded third-party catalogues with the version imported at your site . The page deliberately sits in the system status and not in the settings — transparency that only the administrator can see is none. Search and licence filter help you look things up; the full list can be exported as CSV or JSON for procurement files and auditors. The same listing appears as an appendix in the manual, generated from the same source: two separately maintained lists would drift apart.
• NewNotifications can additionally go through a messenger. Everything reported to you in the application or by email can now also be delivered via Mattermost or any incoming webhook — as a third column in the same matrix you already know: per event and per person. The channel is opt-in without exception; a configured connection sends nothing by itself. Personal matters (assigned, approval, reminder) go as a direct message, team events go into a channel — the two are never swapped, so that no personal notification ends up in front of the whole team. Anyone who cannot be found in the messenger is shown in the administration: otherwise someone believes they are being notified and never is.
• FixedThe discovery collector was never set up during installation — and the wizard reported success. The setup script read the shipped configuration with Windows PowerShell 5.1, which fails on the comment lines in that file. The entire service registration block was therefore never reached: no service, no server address, no enrolment token. The wizard's closing message nevertheless insisted that the service was set up, because it did not evaluate the script's return value at all. Both are fixed: the configuration is now read tolerantly, and the wizard verifies against the result whether the service really exists — otherwise it points to the log instead of congratulating you. The already published collector package of the previous version still carries the faulty script and was deliberately not overwritten; please use the new package.
• FixedThe Windows client crashed when opening the collector overview as soon as it contained an entry. One row of the list referred to a colour that does not exist in the client at all. Such references are only resolved when the row is drawn , not at compile time — and the row was only drawn once at least one collector existed. While the list stayed empty the page could be opened as often as you liked; with the first entry the application terminated on merely opening it. A new guard now checks all of the client's own colour and spacing references against their definitions — in the entire client this was the only hit.
• SecurityThe collector's scan password was in clear text in the setup log. The setup script wrote a transcript, and its header contains the full command line — including the password that was passed in. The logs are located under %ProgramData%\InfraDesk\setup-logs, where every local user may read. The same script properly encrypts this password two lines later. The transcript has been removed; an existing one is deleted on the next run. If you have already set up a collector, please check debug-collector.log in that folder and change the affected password. The password also no longer travels via the command line: the wizard writes it to a temporary file and passes only its path; the helper reads it and deletes it immediately — even if reading fails. An improvement, not perfection, and let it be said: between writing and deleting, the password sits unencrypted on disk. Encrypting it would make no sense here — the helper would have to know the key, and the key would take the same route.
• ImprovedThe token page now states which access does not not come from there. A discovery collector's enrolment token looks confusingly like an integration token — same prefix, same length — and a token issued there can never sign in as a collector, because the required permission cannot be granted to it at all. In pilot operation this cost time exactly once too often. The page now states the difference explicitly and points to the correct route via the collector administration.
• ImprovedRevoked integration tokens disappear from the list instead of cluttering it. Until now every revoked access remained as a row — a list of entries that can never be valid again obscures exactly what the page exists for: which accesses are currently open. Traceability is not lost in the process but moves to where it belongs: issuing and revoking now each write an audit entry — previously neither was audited at all, and the table row was the only trace. Only that entry makes the clean-up defensible.
• NewScan jobs can be removed. What once appeared under Discovery › Scan stayed there for good — the interface knew no delete. Now it does. A running job is rejected rather than deleted: the worker is holding it and writes back its result at the end. The button does not even appear for such jobs — one that predictably fails is worse than none.
• ImprovedThe collector service has an icon again. It was the only one of the four shipped programs without one and appeared in the programs list, in Task Manager and in the services console without a logo — that is, exactly where you look for it.
• ImprovedThe Ordivis Toolbox is now called that everywhere. In a few places in the interface and the manual the old name was still in use.

2026.08.16.140516.08.2026

• FixedThe client could vanish without a word when two dialogs opened in quick succession. This was reproducible when creating a discovery collector. The interface allows only one dialog per window; if a second one opened before the first had finished closing, it took the application down with it. That is the reason for the mandatory update: such a crash leaves no entry in the program log — anyone who runs into it has nothing in hand to report it with. That one place was not the only one, which is why all dialogs in the client now run through a shared gate that queues them one after another instead of opening them at the same time.
• FixedThe comparison against the BSI baseline protection catalogs kept coming back with an error. The BSI restructured its library; the address in the shipped configuration still pointed at the old path and ran into nothing. Correcting the shipped file was not enough: your appsettings.json is left deliberately untouched during an update — otherwise your own settings would be gone after every update — and it would thus have carried the old address forward for good. The retrieval now recognizes the retired path by itself, falls back to the valid source and notes this in the log. A deliberately configured mirror of your own is left alone.
• NewThe legal notices for every third-party component shipped are now enclosed with the product. A file THIRD-PARTY-NOTICES.txt lists the 284 shipped packages with copyright notice, license text and — this is the part one rarely sees — the purpose of use of each component: what Ordivis actually uses it for. The file is installed by all four installation packages; a page inside the application alone does not satisfy the duty to inform. From the same source a machine-readable bill of materials is produced (CycloneDX 1.6, 296 components) with which you can hook the inventory into your own vulnerability screening.
• FixedA component under a restrictive license sat in the delivery without being used. A formula parser under MPL-2.0 came along with the spreadsheet library. The update to the next version takes it out.
• NewA designated reporting channel for vulnerabilities. Under Product security you now find, in three languages, whom to turn to, what belongs in a report and what you can expect afterwards; along with a security.txt at the place provided for it, so that automated tools find the way without having to search. The manual and the in-app help carry the details the Cyber Resilience Act requires from September 2026.
• NewYour license carries a contact address, and every installation reports the version in use. Both serve the same purpose: if a vulnerability becomes known, it must be traceable who is running which version — and how they can be reached. The version travels along with the license check that runs anyway; no new channel arises from this, and a missing entry will never cause an activation to fail. Existing licenses do not carry the contact address yet — it is added the next time one is issued.
• NewIPAM: adopt several addresses in one go. In the address overview several entries can now be selected at once and created together as configuration items instead of one after another.
• NewThe changelog is available in English and French as well. The website is thereby complete in all three languages; the sitemap lists 103 addresses.

2026.08.16.064516.08.2026

• NewVulnerability management — from the manufacturer's advisory to documented closure. Ordivis now keeps your software inventory version by version and reconciles it against two official sources: the CSAF advisories of the manufacturers and the NVD database. A match becomes a Finding with triage, from there where needed a risk in the ISMS, a request and finally evidence that it was closed. The matching runs through CPE 2.3 — the notation in which the catalogues carry their products; deviations in your inventory are maintained as a rule instead of interpreting every advisory by hand. The feed mirror is held tenant-neutral: it is fetched once and read by all tenants.
• NewThe manual is available completely in English and French. All 39 chapters, from the user manual to the database reference — around 41,000 words per language. The diagrams now carry the labels of the respective language as well; a translated chapter with a German diagram in it escapes every text checker, but not the reader. Screen names are looked up, not translated: a set of directions should lead to a menu item that really exists.
• Fixed„Completely translated“ is not what the interface was yet. The previous version reported English and French as finished — and in many places German still stood there: 46 CI class names, the settings navigation, the schedule selection fields, the colour scheme names, eight page titles, the explanatory texts in dialogues and sixteen fixed selection lists. The reason was a checker that structurally overlooked labels; it has been brought up to date, and with it around 900 strings per language. With the selection lists it was checked beforehand whether the chosen value lives on — a translated label that is subsequently stored as a value delivers a different result in every language.
• FixedThe ticket categories appeared almost entirely in German. The existing translations targeted a category list that is not deployed at all — of 130 entries exactly one matched. Now the actually delivered catalogue is covered, all 73 values. When creating a ticket, in the ticket itself and in the bulk change the category therefore appears translated; whoever maintains the catalogue or the rules continues to see the raw value — there an administrator chooses the value that is stored and compared, and they must see it as it is.
• FixedA fallback list in the client showed categories that do not exist in your installation. If the catalogue did not arrive from the server, a hard-wired list stepped in — with 130 entries, of which one matched the delivered inventory. Such a copy cannot be correct in principle, because the catalogue is tenant-specific and maintained by administrators. It has been removed without replacement; instead of a silent fallback list a message now appears. Existing tickets continue to show their category unchanged.
• NewThe wallboard and the Android app now speak three languages as well. Until then the wallboard had none at all localisation at all, the app only German and English. Both had remained invisible because no counter covered them — the language gate measured only the three interfaces that stood in its list. The list now carries all five, and the guard tests look for the language files in the folder themselves instead of enumerating them by name.
• NewThe website is available in English and French. 33 pages per language, including the legal texts — these with a precedence clause: the German version remains authoritative. The TCO calculator formats numbers in the language of the page.

2026.08.15.062715 Aug 2026

• NewThe interface is now available in French as well — in full. Client, web portal and the messages of the interface are translated; that amounts to 5,284 strings. The language is chosen by each user individually, not by the tenant: in the client under Settings › Appearance, in the portal through the flag icon. In the client the switch takes effect after a restart; the portal changes immediately. The names of the languages are given in the language itself — someone who cannot currently read the interface will find „Français", but not „French". Your own data remains untouched: Ticket subjects, CI names and notes remain in the language in which they were entered; the interface changes, the content does not. If a translation is ever missing, the German text appears — nothing is left blank.
• FixedThe web portal would not have accepted French at all. Internally it listed only German and English as permitted languages. A language missing there is silently reset to the default language — without an error message and without a log entry; the page simply stays German. The finished translation file would have been unreachable. Two checks now catch this silent slip: a translation that exists must also be selectable, both in the portal and and in the client.
• FixedThe note beneath the language selection was out of date. It claimed that switching was disabled — while standing right beside the selection list, which had long since worked. It now says what applies: switching takes effect after a restart, and missing translations appear in German.

2026.08.15.003515 Aug 2026

• NewSeveral separate Ordivis installations from one client. Not every organisation runs everything on one server. Anyone operating the core administration and the municipal utilities separately had to rewrite the server address by hand before every switch. The administrator now records the installations once as a named list; a selection field appears above the sign-in form — visible only from two entries onwards. If the list is absent, nothing changes: existing installations carry on unchanged and the selection stays invisible. Not to be confused with multi-tenancy — there, several tenants share one installation; here one client reaches several separate ones. The Android app receives the selection too, on managed devices through the MDM solution; the web portal gets a list of links to the other portals.
• ImprovedThe certificate option and Windows sign-in now belong to the individual connection. Both applied to the entire workstation until now. With one server holding a valid certificate and another a self-signed one, that was simply wrong: switching verification off for the test server switched it off for the production server too. Likewise, Windows sign-in can now be enabled for the production server and not for the test server.
• SecurityThe stored Windows sign-in is now kept separately per connection. There was exactly one store. After a switch, a refresh token would therefore have gone from server A to server B — to a server that never issued it. The store now carries an identifier of the server address; the existing one is taken over once at first start so that nobody has to sign in again. The list itself is protected against write access by ordinary users, and the setup reads the permissions back instead of merely setting them — on a data drive the branch does not inherit them by itself.
• ImprovedWith several installations, a lead profile determines which build a workstation is raised to. Without that decision it would move to the version of the newest server it had ever connected to, and would then speak to an older interface it does not know. The update service now follows the lead profile instead of the connection last chosen, and the out-of-date notice in the client offers the upgrade only there; for the other connections the notice appears without a button.
• NewThe BSI's ready-made justifications can be imported — until now Ordivis only produced this layer itself. Alongside the catalogue, the BSI publishes component definitions: for Keycloak, network architecture, password policy, supply chain security and the GA-Lotse base module it already states which requirements they implement and how how. Anyone operating one of these components now finds the justification already written out instead of having to write it. Only what is confirmed is adopted — the BSI states expressly that these definitions are references and do not replace the action of the responsible body. Until confirmed, every entry visibly carries „suggestion".
• NewISO 27001 and BSI Grundschutz++ can be read against each other. Anyone running both — the normal case in public administration as soon as a certification is added — maintained every implementation twice. The official mappings between ISO 27001 Annex A, the former IT-Grundschutz and Grundschutz++ are now imported. The type of relationship stays visible and is not compressed into a tick: only „equivalent" carries an implementation across unchanged, and in the ISO collection that is just 16 of 96 mappings. Whatever points to a different catalogue version is counted and displayed rather than omitted.
• FixedFour citations of state law were wrong — one of them named an act that does not exist. Thuringia is the only federal state without an act of its own on the public health service; a regulation still applies there. What was stated was an act that in August 2026 is still going through the legislative process. Also corrected: the Saxon local elections act, the data protection implementing act in Saxony-Anhalt and the emergency services act in Schleswig-Holstein. That brings 20 of the 31 areas of law checked against a source; the remaining eleven deliberately cite only the Hessian reference of the template and assert nothing for any other state.
• FixedAn unreadable mapping file produced a process catalogue that presented Hessian state law as valid nationwide. In the column that marks state-dependent citations, an empty value means „federal law, applies everywhere" — not „unknown". A first run without the abbreviations loaded therefore created 5,019 processes, of which 1,318 carried a false statement; it was only healed at the next start. Now nothing at all is created in that case: a wrong legal basis is worse than a missing one.
• FixedOne label stood in German in the English program. The enumeration of process types was affected. From now on a guard reports every English text that still carries German umlauts — that is the only remainder that can be demonstrated mechanically.

2026.08.14.171514 Aug 2026

• SecurityFour tables of the monitoring integration were the only ones without a tenant filter. Affected are exclusively installations in which two tenants share one database — as a rule each tenant has its own, but not always: legacy tenants from a migration, as well as archived and quarantined tenants, fall back to the shared connection. There, source and alert lists named hosts and services of the respective other tenant; a monitoring source belonging to one could be renamed or switched off by the other, and a suppression rule without a source binding swallowed that tenant's tickets along with it. The rows carried the correct tenant id and therefore escaped every check. The filter now applies to all four tables. Anyone running a separate database per tenant — the default — was not affected.
• NewThe raw monitoring events are pruned after 30 days. Until now they grew with the volume of events rather than the volume of tickets: a monitoring system with a thousand checks writes tens of thousands of rows a day in a storm. A nightly job now clears them out. The episodes — that is, the history of the disruptions — remain untouched, and events of a disruption still open or in its grace period also remain: deleting the wording for a running case of all things would take away your evidence towards the operator of the monitoring system. Timing and activation are under Administration → Jobs.
• ImprovedOn request the warning map shows only one type of warning — and can be enlarged. A nationwide heat warning colours the map across the whole area; beneath it an approaching thunderstorm was no longer visible — and that is exactly what this page is for. Selectable are UV, heat, thunderstorm, wind, rain, snow, black ice/frost, thaw and fog. The lightning is unaffected by this and can be switched on alongside the thunderstorm. The codes come from the DWD's own presentation specification, so they are not guessed.
• FixedOn the warning map, state borders and the rain radar lay beneath the warning areas. An opaque warning area therefore hid both — anyone switching on the radar saw nothing and concluded that it did not work. And without recognisable borders, one's own area could not be found on the map.
• ImprovedIn the web portal, Processing and the satisfaction question are now in the side column. Both sat beneath the journal in the wide column: anyone wanting to change the status had to scroll past the description, the history and the reply field. In the Windows client it was always on the right — now it is here too. On a phone and on a tablet in portrait the blocks stack beneath each other as before.
• FixedSix labels of the warning map were missing in English. Affected were warning type, image height, lightning and storm cells, radius and the two zoom buttons. A missing translation does not fail — it silently shows English-speaking users the German text.

2026.08.14.125514 Aug 2026

• NewOfficial severe weather warnings from the Deutscher Wetterdienst — down to the municipality. Until now the weather display hung on a weather station. But the DWD does not warn for stations, it warns for warning cells, that is for municipalities — a station in the neighbouring district could report calm weather while a severe weather warning was in force for your own town. Warnings now come from the official source for the tenant's municipality, with warning level, period and the official wording; along with the DWD warning map and the forest and grassland fire index. Visible in the web portal, in the Windows client and on the wallboard. Anyone holding the „weather warning" role is notified of a new warning. The DWD's test messages are filtered out — apart from a single field they look exactly like real warnings.
• NewThe service desk can now be configured in the web portal, not only used. Until now ticket categories were writable there and SLA profiles readable; everything beyond that needed the Windows client. New in the portal: SLA profiles with response and resolution deadlines per priority, service-hour calendars with time windows and public holidays per federal state, escalation levels, the e-mail signature per support group with a preview, the rule set of routing rules, automation rules and maintenance windows, support groups with OLA targets and memberships, text snippets and ticket templates, the data quality overview and a situation overview in large figures. For support groups, the rules and the people that follow from them stand side by side — otherwise it remains unexplained why someone nobody entered sees the group's tickets.
• NewThe organisation modules are now usable for companies as well. New are a process map over your own processes, a supplied reference map with 52 main processes under ISO 9001, a CSV import for your own frameworks, the establishment plan with target against actual, job descriptions under ISO 9001 clause 5.3, authorisations per post with an amount limit and a four-eyes requirement, the responsibility matrix, the organisation manual as a frozen version with a checksum, and process maturity under ISO/IEC 33020. The establishment plan expressly does not not count deputies as filling a post: a deputy steps in when the post is vacant — they do not occupy it.
• FixedA corporate tenant saw 5,019 public-administration processes under „process organisation". The page loaded exclusively the public-administration reference catalogue, whose entries are visible across tenants. A company was therefore shown a municipal process map complete with the filter „municipality/city, municipal association, state", while its own maintained processes stayed invisible. Catalogue tree and administrative tier now appear only for tenants in the public administration sector; in their place, companies get a search over process name and business capability.
• FixedDeleted SLA profiles and service hours came back after every update. Reported from pilot operation, where the shipped profiles are deliberately deleted. The shipping seeder asked „is there any SLA profile at all?" and concluded from that whether it had already run — the two are not the same thing. It runs at every start of the service for every tenant; anyone deleting the shipped data made the check false again and had the data back after the next restart. A deliberate deletion was therefore impossible. The seeder now records that it has run rather than guessing it from the presence of its own data. The same affected the service catalogue, whose items fell back to the shipped state at every start, and the main categories. Anyone who has already tidied up gets nothing back with this update — a back-fill during installation sees to that.
• FixedThe automatic update found a release only a day later. The nightly maintenance window read only a stored value that a separate check refreshed during the day — discovery therefore lay six hours behind the period in which anything may be applied at all. A release published at 22:15 was not seen at 04:00, was found at 10:00 and was applied only at 04:00 the following day; nowhere was there an error. The path that applies the update now asks for itself. If the operator has switched update checking off, it stays off; an unreachable update server does not cost the window. This takes effect from this version onwards — the running one does not carry it yet.
• FixedThe weather page showed Berlin to every tenant. The weather station was fixed in the configuration file and carried the same value in every installation. For a municipality this page is the way to the official severe weather warnings — a warning for Berlin is worse than none at all in the Fulda district. The location now follows from the postcode in the tenant profile; if it is missing, the page says explicitly that the location is a guess. The key Weather:StationId is no longer read — a deliberate setting now lives under Weather:StationIdOverride.
• FixedThe connection test of the Exchange integration failed with „the specified time zone is invalid". The time zone transmitted carried two identical transition points, and no usable time zone can be built from that. Not only the test was affected but every free/busy query.
• FixedThe source of the BSI Grundschutz++ catalogue was dead — and the daily check still reported success. The BSI has restructured its library; the previous path no longer answers. A permanently wrong path looked like a brief network glitch, anew every day, without the job ever turning red; the installation believed itself current because nobody contradicted it. Failures are now told apart by kind: timeouts, network errors and broken data remain a warning, an address that no longer exists stands out. The sample catalogue supplied is at the state of 13 Aug 2026 (652 instead of 643 controls).
• FixedA company chose the type of its unit from „department", „office" and „section". The list of unit types was maintained independently in four places, and the four contradicted each other — a value the interface offered could fail validation. There is now one source with a sector mapping, extended by the missing business terms company, executive management, business division, plant and branch. What is filtered is only the selection, not validity: anyone changing sector or importing existing data keeps their values.
• NewThe mailbox now has a say in category and responsibility. An enquiry to the sales address was until now collected from the right mailbox and still landed as an incident without a category and without a group — indistinguishable from a support report. The inbound path now evaluates the rule set, the mailbox form offers ticket type and priority, and permitted main categories and a default category can be set per mailbox. A rule cannot break out of that. Existing mailboxes behave unchanged.
• NewA group view: one ticket list across several tenants. Whoever holds the new permission sees their own tenant and every further one for which a valid grant exists — in one list, in the same order as before. Without a single grant the view delivers exactly your own tenant. If a tenant's database does not answer, it is named explicitly instead of silently missing.
• NewAn attached e-mail can be read in the portal, not merely downloaded. Anyone without an Outlook workstation — in the self-service portal the majority — until now held a file they cannot open, while the images from the same mail stood visibly in the history. Attachments inside the mail are named but not delivered: they never passed the inbound check. If the message came from the reporter of the ticket, they now see their own mail in the history too.
• NewAn overview „which state law applies to us?" The relevant state acts on data protection, information security, e-government and archiving, derived from the tenant's federal state.
• ImprovedThe ticket list in the Windows client keeps itself up to date. On a new ticket or a reply, only the notification at the bell appeared until now; the list showed the ticket only after navigating away and back. If rows are selected, no reload happens — otherwise a bulk change in progress would lose its selection.
• ImprovedIn the portal's ticket overview the filters take effect immediately. Status and priority took effect only at the next keystroke in the search field. New is the combined value „everything except closed".

2026.08.13.211613 Aug 2026

• FixedAfter a fresh installation no backup ran — and the interface still showed „daily 02:00". The scheduled task was created in exactly one place: on saving under Settings → Backup. Anyone who never went there after setup had none. The display reflected the stored intention and not the task scheduler — it looked configured and was not. The installer now creates the task itself, the service checks it at every start and creates it if needed, and the interface states explicitly when no task is registered. After the update, please check the settings for a warning — on existing installations the task creates itself at the next service start.
• FixedAn image from an e-mail did not appear in the history where it belonged. Reported by a reporter about their own ticket: in the web portal a gap remained where the image should have been, and the same image stood at the very bottom of the attachment list — with no visible connection to the message it came with. Whoever sees that concludes their attachment did not arrive. Two causes: the image address was aimed at the Windows client and could not be retrieved in a browser; and attachments hung on the ticket, not on the entry that brought them. Both are fixed. Whatever came with a message now stands with it — in the client as in the portal — and does not appear a second time in the collective list. For tickets from before that time the connection can no longer be reconstructed; their attachments remain at the bottom, where they stood before.
• NewA ticket now refreshes itself. Until now you had to leave the page and come back to see a new reply or a change of status. As long as a ticket is open, the server reports every change to it and the view reloads by itself — in the Windows client as in the web portal. For now in the ticket view, where it shows up daily; further pages will follow.
• NewImages can be copied from the preview to the clipboard. A screenshot from a ticket can thus be taken into a mail, a document or another ticket without the detour of downloading and re-attaching. In the client from the image view, in the portal from the preview in the history.
• NewThe client and the wallboard no longer need a separate Windows runtime. Both now bring the interface runtime with them. The installation step that previously required administrator rights and had a failure mode of its own disappears entirely — the wallboard therefore installs completely without a prompt. The packages even become smaller: the wallboard setup from 88 to 44 MB, the workstation setup from 148 to 104 MB.
• NewA separate, lean package for the workstation — as a setup and and as an MSI. The workstation was until now available only through the full installer (~665 MB), which brings server, portal and database parts that have no place in a workstation network. The new package contains only the client and its update service. Additionally as an MSI, so that workstations can also be distributed by group policy — with a setup file alone that route stayed shut. Both accept the server address unattended (/SERVERURL= or SERVERURL=) and abort explicitly without it, instead of silently falling back to a default. For updating existing workstations the update service remains the right route, not an installer.
• FixedThe software inventory showed the wrong manufacturer and an out-of-date version. Both entries come from the Windows record under „apps & features", and only the installer writes it — an update exchanges files and leaves it untouched. It therefore aged further with every update, and an inventory reported for months a version that was no longer running anywhere. The record is now corrected to the version actually running at every service start; the manufacturer reads „Grams IT" as intended. This also takes effect on existing installations, without a reinstall.
• FixedThe notice of a new version reached only one tenant. In an installation with several tenants the message for all every recipient was stored in the database of the first. Nobody saw it there: the recipient looks in their own. The administrators of the remaining tenants therefore learned nothing at all about a mandatory update. Every message now lands with its recipient's tenant.
• NewRack mountings can be exported as a table and read back in. Anyone clearing a rack or transferring a stocktake had to set every rack unit individually. The export delivers all mountings of one rack or all racks, the import takes them back — by default as a dry run that only shows what would happen, with a finding per row. Only the second pass writes.
• NewRack markers for photography. The position and edge length of a marker stuck to the rack can be recorded — the basis for deriving measurements from a photo. The manual describes where and how the marker should be attached.
• FixedThe installer's connection check failed on its own HTTPS redirect. It reported an error although the service was running perfectly. The check runs through the Windows PowerShell, and that cannot reach the redirected endpoint. The health query now answers without a redirect — exclusively over the loopback, so only from the machine itself.

2026.08.13.145813 Aug 2026

• FixedThe backup contained only the database — not the key that makes its contents readable. Credentials for the directory service and the mail server sit encrypted in the database; the key for them lives in a file beside it that was not included in the backup. Anyone restoring such a backup got a running installation with unreadable credentials — directory synchronisation and mail sending would have fallen silent, with no visible error. The backup now additionally contains this key, the operating settings and, where configured, the TLS certificate. This happens only in the encrypted backup: placing the key unencrypted beside an unencrypted database would mean giving both away together. Without a backup password on file it therefore stays at the database alone — and the log says explicitly that this backup is incomplete for a restore. Anyone who has not yet set a password should do so now. Backups that already exist do not contain the addition; the restore reports that too, instead of assuming completeness.
• FixedThe restore wrote no log and left the services stopped after an abort. Found during its first complete test under real conditions. It deliberately runs decoupled in the background because it stops the very service that would otherwise end it — there is no display there, and so it left no trace: neither of its course nor of an abort. If it aborted, the installation was moreover left with the service stopped; a failed attempt therefore ended worse than none at all. There is now a log beside that of the backup, and the services come back even after an abort — the data remaining untouched in the process.
• ImprovedThe restore reports unambiguously whether it succeeded. Until now it surfaced a technical side message from the database tool that also occurs in the normal case — success and failure could not be told apart from it. What counts now is whether the services are running again.
• FixedThe backup scripts lay in two places, one of them out of date. The second location was filled once during setup and then stayed put, while the version actually used is carried along with every update — and the restore script was missing there entirely. Anyone reaching for it in an emergency gets old behaviour; an out-of-date copy of a restore script is more dangerous than none. Both now live in exactly one place.

2026.08.13.070813 Aug 2026

• FixedAn update started from within the program aborted after 30 seconds and was rolled back. The entire run hung on the time limit of that one call. If the server did not answer within 30 seconds — and during an update it never does, because backing up, downloading and swapping all happen — the program dropped the connection and the server aborted the run in the middle of downloading. An update whose files take longer than the caller waits could never succeed by this route: every larger delta, every slower line. Small updates stayed below the limit and went through — which is why it went unnoticed for a long time. The run no longer hangs on the call: it carries on even if you close the window, and aborts only when the service is genuinely shut down. Aborting an update in the middle of the file swap would in any case be more dangerous than letting it finish. For installing this version: the fix takes effect only after it has been installed — the update itself is still carried out by your previous version. If you are coming from an older version with a larger delta ahead of you, please therefore do not not use „update now" but „schedule" or the fully automatic update in the maintenance window: these routes run in the background service and do not know the time limit. The setup file is equally reliable.
• FixedThe error message pointed at the wrong place. The log said that at least one file was missing on the update server or that its checksum did not match — while the file named lay there perfectly. Anyone following that up searched in a place where everything was in order. The cause was that an abort and a missing file could not be told apart internally. An abort is now reported as an abort and expressly not as a finding about the delivery; the installed files remain untouched in the process, and a fresh run starts cleanly from the beginning.

2026.08.13.051313 Aug 2026

• FixedAn incoming message was deleted from the server even when no ticket arose from it. Since the previous version the mailbox poll clears up after itself — properly safeguarded for the ordinary route, because there the complete original mail with all headers sits on the ticket. One special case slipped through, however: if a module takes the message — the backup module, for instance, which reconciles a report mail against a backup run — then no ticket arises and therefore no no archived original either. It was deleted all the same. That would have thrown away exactly what one could still have fallen back on if the module misread it. What is deleted now is exclusively what demonstrably exists as a ticket with its original; messages taken by a module remain in the inbox as read, with a note in the log.
• ImprovedThe directory synchronisation reports the change instead of the stock. The previous version's warning named the number of users without a role — and therefore appeared on every run: in an administration with many service and function accounts, several hundred are permanently and entirely rightly without a role. A warning that always comes is no longer read after a short while, and the one run on which something really goes wrong is lost in it: if a deleted group assignment cost forty people their permissions, only a digit in that line would change. What is reported now is an increase compared with the last run — and additionally an assigned AD group that is no longer found on any user because it has been renamed, moved or deleted. The plain stock is still shown by the directory view, where one looks on purpose.
• FixedThe operations manual named a script that was not shipped at all. Section 5.1.2 described it as the route for nudging an existing workstation once — but all of the installer's helper scripts were deleted after installation, so it lay on no installation at all. They now remain in the subdirectory scripts of your server installation: both the script that raises the update service of an existing workstation and the one that sets it up where it is still missing. The source is the directory clientupdate of the same installation — you do not have to provide anything additionally. The manual section now names the actual paths and distinguishes the three possible states of a workstation.
• ImprovedManual: two changes in behaviour added. Newly described is what happens to an e-mail after processing (ID-31, section 3.1) — anyone who would otherwise have noticed an emptying inbox finds the explanation there. And how AD groups become roles (ID-41, section 8.2): that membership is resolved across all levels of nesting, that the run only grants roles and never revokes them, and that the number of assignments reported counts only the new ones.
• ImprovedA library for encrypted connections was raised to the patched version. The vulnerability reported concerned a part of the library this application does not use; it was raised nonetheless instead of silencing the report — a suppressed warning is noticed by nobody at the next, genuine finding.

2026.08.12.181512 Aug 2026

• FixedThe directory synchronisation saw only the groups in which a user is directly a member. In an administration that has grown over time that is the exception: what is assigned is a department's group, and it contains not the people but the groups of the individual services. At a pilot installation the department group held four people and four further groups — everyone below them was left out. The result was the same as before the previous version: 284 of 351 accounts without any role, able to sign in and unable to act afterwards, with the assignment maintained completely and correctly. Membership is now resolved across all levels of nesting — by the directory server itself, which keeps this chain anyway. Direct membership continues to be evaluated in addition: if nested resolution fails for a single group — with a directory service that does not know this query, for instance — the previous route carries it along, and the remaining groups are unaffected. After the update one synchronisation run is needed so that the roles are filled in; that does not happen retroactively by itself.
• FixedA comma in a group name would have caused the query for that one group to be rejected. Groups named after a person carry one — and different special characters apply in the search filter than in the name itself. The failure would have affected only part of the assignments and would therefore have been particularly hard to notice: everything else would have carried on working.

2026.08.12.163012 Aug 2026

• FixedThe directory synchronisation created users but assigned them no role at all. Not even when the mapping from AD group to role was maintained completely. Those affected could sign in and then do nothing at all: every call was refused, and in the self-service portal „my tickets" reported only that nothing could be loaded. In an administration with more than 60 accounts, nobody was able to act — and the synchronisation reported every run as a success. The cause was a missing item in the query: Active Directory returns group membership only if it is requested explicitly since it is one of the fields computed only at runtime and not included in a query for „all fields". The synchronisation therefore saw a group on no user at all. After the update one synchronisation run is needed so that the roles are filled in. In addition the run now reports explicitly when it creates users and cannot assign a role in the process — that number previously appeared only in passing in the log.
• FixedReplies from the agent did not reach the reporter when both use the same mailbox. In smaller organisations that is the normal case — a shared info@ or accounts@. The distribution list excluded the sender by address rather than by person and thereby removed the reporter as well; nobody was left, and sending ended silently. That the resolution mail still arrived on closure — it takes a route of its own — made the failure look like a coincidence. The exclusion now applies to the person; if a different account sits behind the same address, it stays in the list. A distribution list that ends up containing nobody is logged instead of being passed over silently.
• ImprovedThe fault report now goes directly to the vendor instead of through your mailbox. Until now it went out as an e-mail and therefore assumed that your mail server may deliver to an external domain. An Exchange does not do that in its default configuration; at a pilot installation this caused three reports to fail for good, while the sender read „report sent" every time. The report now takes first the secured route over which your installation fetches its updates and checks its licence anyway; e-mail remains as a fallback for installations without internet access. For this, no secret is stored on your installation. The program now also distinguishes between „delivered" and „queued" instead of reporting both as sent.
• NewAn e-mail that is permanently undeliverable reports itself to the administrators. After six vain attempts a message counts as lost. Until now that stood only in a log file nobody looks at as long as they have no suspicion — which is exactly how the three lost fault reports went unnoticed for days. The message names recipient, subject and the mail server's own wording, so the diagnosis comes with it.
• NewA ticket's history now shows which entry went out. A small envelope icon on the right-hand edge that names the recipients on hover. Until now an entry nobody received looked exactly like a delivered one — „written" and „sent" could not be told apart. For entries from before that time the icon stays off: guessing after the fact who received a mail would assert a delivery that was never verified.
• FixedA fault report from a demonstration tenant vanished without trace. The block that keeps demonstration data from sending mail also caught the report to the vendor — although that goes to a real address and cannot produce a non-delivery report. No entry appeared in the queue, no note in the log, and the sender still received a confirmation. The report and the support package are now exempt.
• ImprovedEvery log file says in its first line where it lives and where that information comes from. From the central setting or from the default — that answers the first question in an incident without looking up three configuration sources. If the folder cannot be created, the reason now appears in the Windows event log instead of only on a console that a service does not possess: until now the file log was off in exactly that case and and the reason for it was untraceable.
• ImprovedThe mailbox poll clears up after itself. Processed messages stayed in the inbox and were merely marked as read — it grew without limit, and anyone looking in could not tell what the system had already turned into a ticket. They are now deleted as soon as the ticket has been created; its content then sits on the ticket. Anyone wishing to keep the messages enters an archive folder in the mailbox settings — they are then moved there instead of deleted. What could not not be processed goes to the error folder unchanged and remains findable.
• FixedThe mailbox poll reported „no mailbox configured" and „poll running" in the log at the same time. Both statements were true — only for different tenants. A shared flag flipped back and forth on every pass and produced a contradictory pair every 90 seconds that looked like a flapping outage. The message now names the tenant and applies only to it.
• ImprovedThe operations manual describes in section 5.1.2 the one-off nudge for workstations already set up. Anyone reading only the manual would otherwise never learn why their update service stays on an old build while it continues to raise the workstation client reliably.

2026.08.12.112712 Aug 2026

• FixedThe update centre reported „up to date" as soon as the API was up to date — the portal, the workstation client and the wallboard were not asked. That is one of three reasons for marking this a mandatory update. If one of these components was behind, the change plan was not even computed: the update was unobtainable through the interface. The lag was determinable the whole time — it simply was never used before the display. Every installed component is now checked individually, and the update centre names the one that is behind. A component with no determinable version is deliberately treated not as behind: an absence of information is not a finding.
• FixedThe workstation client's fallback route was unusable on exactly the machines it exists for. The helper program that performs the file swap was shipped without a runtime of its own. On a plain workstation, however, the setup installs no .NET runtime — the client brings its own — and there it did not start. On a machine with a runtime installed, something like that never shows up. Both helper programs are now shipped with their own runtime, and the build checks this against the runtime identifier instead of the return value: shipped and able to start are two different things.
• NewFrom this version onwards the workstation's update service keeps itself up to date. It was the only shipped component without an update route of its own: it kept the whole workstation current and itself stayed at the level of its installation. It is now part of the delivery package and swaps itself — without a prompt, since a background service interrupts nobody — before the component it maintains. Existing workstations need exactly one nudge from outside for this, because a capability cannot deliver itself; for that, deploy\update-clientupdate.ps1 is included, which touches this service alone — no database, no program data. After that it runs by itself permanently.
• ImprovedThe operations manual now describes the one-off nudge for workstations already set up. Anyone reading only the manual learned nothing of it — the update service would have stayed there on an old build while continuing to raise the workstation client reliably; from the outside that looks unremarkable. Section 5.1.2 now names the call, what the supplied script touches (only this service's directory — no database, no program data), that it requires administrator rights and, without them, aborts before before the first change, how the run can be previewed with -WhatIf and how to tell that it succeeded: after the swap the version is read from the program itself, not from the return value. Per workstation the step is needed exactly once; after that the update runs by itself permanently.
• FixedIn the installer, consent once given to removing „old remnants" could survive a change of installation type. Anyone choosing „server", confirming the prompt, then going back and switching to „workstation" would have triggered the clean-up on a plain workstation — and with it the database, the backups and the key. The consent is now re-checked at the point of use against the current installation type; the change can only prevent the clean-up, never trigger it. When updating, the answer to this prompt remains „no".
• ImprovedThe web portal signs out after 30 minutes of inactivity instead of eight hours. Please read before installing: from the update onwards this applies to existing installations too, without anyone changing a setting. The large gap from the API is deliberate — at the Windows client a named person works at a workstation they lock; the portal is open to end users and runs on shared machines too. The period is sliding; every access resets it. Anyone wishing to keep the previous value enters, in the appsettings.json of the portal, Session: LifetimeMinutes: 480 as the setting.
• ImprovedThe session lifetime of the Windows client rises from eight to twelve hours. It too is an idle period and not a maximum: every renewal resets it. Twelve hours cover a working day including a break and a short sleep. In the process it emerged that the installers had always set a key for this that does not exist under that name — the value was silently discarded at startup and the defaults applied throughout. On every installed server there was therefore a setting that looked like an intention and never took effect. The name is corrected, the old spelling is removed from the environment on installation, update and uninstallation, and a guard now checks the installers' key names against the code.
• ImprovedThe ticket number and reply token now stand at the end of the mail subject. As a prefix they took up the entire visible subject line on a phone — in the mail list all ticket mails looked the same. The matching of incoming replies hangs on exactly this marker; it is therefore now recognised as a unit, and specifically the last one in the subject, so that neither „Re:" prefixes nor a forwarded history displaces it. Existing mails with the marker at the beginning are matched unchanged, and replies to them do not break.
• FixedThe bug button in the workstation client did not open when a dialogue was already open — that is, precisely when something had gone wrong, because whatever goes wrong the client shows as a dialogue. No report was written, and nothing further happened on screen. The report now clears the way for itself; the text of the open message appears in the report anyway as the last error shown.
• FixedSystem messages went out through another tenant's mailbox. With an application-wide trigger — the notice of a new version, for instance, which reaches the administrators of all tenants — no tenant was carried along, and the mailbox choice reached across tenants. Every recipient now receives the message from their own organisation, and their reply does not land in another customer's mailbox. If a tenant has no mailbox of its own, the configured fallback applies — a visible failed attempt rather than a silent wrong sender.
• FixedDemonstration tenants no longer send e-mail. Their addresses use the suffix reserved for examples, .example for which no mail server can exist — every message bounced back into the sender's mailbox as a non-delivery report. Blocking happens at queueing, not at delivery, so that a message does not end up as „failed" as though something were broken. The in-app channel stays open: that is exactly what a demonstration is meant to show, and it does not leave the building.

2026.08.12.055612 Aug 2026

• FixedAttachments lay in the service account's temp directory — and were therefore covered by no backup. That is the reason for marking this a mandatory update, and it is a fault we found in our own work. The storage location was never explicitly set: no shipped configuration contained a path, so a fallback to the temp directory applied every time. Windows may empty such a directory at any moment; the daily backup is a database dump and does not cover it; and if the service later runs under a different account, the path moves with it. The record in the database survived all of that every time — the content did not. An attachment could disappear without a message appearing anywhere: it still stood in the list, and on opening it was gone. From this version onwards, attachments therefore live in the database. A single transaction writes record and content together or not at all, the attachment sits in the dump of your daily backup, and a restored database brings its attachments with it. A second backup route for a file directory becomes unnecessary.
• NewThe existing stock is migrated automatically at first start — once per database. The run keeps four promises. The previous location is forgotten only once the content has actually been written is; there is no intermediate state. A file that cannot be found is reported in the log rather than passed over, and its record keeps the old path so that a later run can catch up with it. Nothing is deleted — the old files stay where they are until you have checked the migration. And the run never aborts startup: a single attachment that sticks is an operational problem and no reason not to bring the platform up. The search covers every place it could have been, including the system account's temp directory — so the stock is found even if the service now runs under a different account.
• ImprovedThe operations manual describes the migration in the section Backup & restore. With the log lines the run writes, the locations searched, a query for what remains — and how to supply a different directory should the old files now lie elsewhere.
• FixedThe chapter on the monitoring integration could not be found through any index in the manual set. It was missing from the document map and carried no label on its title page — a faultless chapter to which no route led. Both are fixed. The database reference in the manual has also been regenerated and now knows the tables of the monitoring integration and the extended IPAM fields.

2026.08.12.013012 Aug 2026

• NewMonitoring integration: alerts from a monitoring system become tickets. Deduplicated by the disruption episode, so that fifty alerts about the same disk produce one ticket with a counter instead of fifty tickets. Connection is by webhook or through a mailbox; both take the same route.
• NewAll-clear with a grace period. The ticket resolves itself as soon as the monitoring system gives the all-clear and the period has elapsed. If the disruption returns before then, it stays the same ticket. If work has already been done on it, only a comment is added — a monitoring system does not know whether someone is still checking.
• NewMaintenance windows suppress ticket creation but keep the alert on record. In a storm, further alerts hang on a collective ticket instead of flooding the mailbox.
• NewRules on host, service, severity and labels. It determines category, support group, assignee and ticket type — or that no ticket arises at all.
• NewAn alert overview with the last receipt per source. A connection that has delivered nothing for days otherwise looks like calm. Recurring disruptions with the same key are shown separately: every individual ticket was properly resolved; the pattern only becomes visible when they are seen together.
• NewIncluded: a notification script for CheckMK and instructions for the PRTG Execute HTTP Action.
• NewA rack occupancy plan in the workstation client. Front and rear view with a rack-unit grid, occupancy, free slots including the largest contiguous gap, and a collision warning while entering data. Entry point from the rack.
• NewA calendar with weekday names and calendar weeks. The start of the week can be set to Monday or Sunday per user.
• NewA whole address range can be assigned to a network interface — a public /28 network on a firewall's WAN port, for instance.
• NewThe deputy can be given directly when creating and editing a post. Anyone who did not find the previous second step created a post of its own called „dep." instead — and the two knew nothing of each other.
• NewThe wallboard can be pinned to a particular screen and survives an error in the interface.
• FixedA maintenance window in the calendar led nowhere when clicked and was therefore not findable. Maintenance windows and change requests now jump to their page and are highlighted there.
• FixedA device's interface list showed only the address read most recently, although several could be assigned.
• FixedAn empty entry in the e-mail signature left its punctuation behind. A municipality with no legal form got empty brackets after the company name.
• ImprovedIn the web portal the notification settings live in exactly one place instead of two.
• ImprovedThe backup module is fully translated.
• ImprovedEvery fifteen minutes the wallboard records that it is running, and notes an orderly shutdown. If the log previously ended in the evening, there was no telling why the screen was black in the morning.

2026.08.11.195011 Aug 2026

• FixedA single long e-mail could bring the entire inbox to a standstill. That is the reason for marking this a mandatory update. If a message arrived whose text was longer than a ticket's description field holds, processing broke off — and because the mail was therefore never marked as read, the next poll fetched the same message again and broke off again. Everything behind it in the mailbox never arrived: no new tickets, no replies from reporters, no indication in the interface. From that moment reception stood still until someone cleared the message out of the way by hand. A forwarded conversation or an ordinary HTML mail is enough for this. Long texts are now shortened to the field width — with a visible notice instead of silent truncation; the full wording is retained in the history and in the archived original mail. The same applied to long replies to existing tickets.
• ImprovedA message that cannot be processed no longer holds up the rest. It moves to the error folder, which had always been configurable on the mailbox and was used nowhere until now; if the folder cannot be created, it is marked as read and stays in the inbox. Either way it remains findable, and the poll carries on. Nothing is lost in the process — with POP3 such a message is expressly not not deleted, because there it would be irreversible. How many messages were affected is recorded in the log, with sender, subject and reason.
• Fixed„Test connection" reported „all in order" and deleted in the process the very error one was looking for. The test establishes the connection and signs in — no more. It says nothing about whether a message also becomes a ticket, yet on success it cleared away the poll's error message and error counter. Anyone investigating a reported error and testing while doing so deleted the finding; at the next poll it reappeared, and the next test deleted it again. The display therefore looked erratic and the error looked like a coincidence. A passed test now proves only reachability — visible from the time of the last success.
• ImprovedThe mailbox now states what what went wrong. Until now the same sentence stood there for every kind of error: „An error occurred while saving the entity changes. See the inner exception for details." That is the outermost layer of an error message and does not name the reason — that lay one level deeper and was discarded. The actual cause now stands there.

2026.08.10.223510 Aug 2026

• NewBackup: plan, record and demonstrably check. InfraDesk does not back up itself — it steers the backup products already in place. For each backup plan you record what has been promised (protection class, owners, deputy, RPO and RTO), and for each job what the product actually does. The schedule is a construction kit rather than a free-text field: daily, on particular weekdays, on a calendar day or on the last Friday of the month — with a start window, a time zone and a tolerance. A preview of the next ten occurrences shows before saving whether the plan means what you have built; across the summer time change as well.
• NewThe run that did not happen becomes visible — and that is the real gain. A backup program whose service no longer starts reports nothing: no error message, no mail, no line. In a system that only collects messages, this state looks exactly like „no problems". InfraDesk therefore knows the expected run and records it as missed if nothing arrives by the end of the tolerance window. A late report lifts the record again; an announced maintenance window produces a note instead of a false alarm.
• NewSign-off in one click — but only if the period allows it. Review periods are created in advance from the plan's rhythm, with a deadline and an owner. If the period is green throughout, you close it in two seconds. If a night is missing or a run is not green, the button stays disabled and names the reason — the route then leads through a finding for each affected run. What is stored is not a tick but evidence: the number and identifiers of the runs, the volume backed up, the time and the reviewer. A later correction places a new version alongside instead of overwriting the old one.
• NewA failed run becomes a ticket at the touch of a button. The subject is composed from job, system, affected device, time and error message; the configuration item is linked, and it is assigned to the backup owner, failing that the deputy, failing that the support group. Fourteen failed nights in a row produce one ticket with fourteen journal entries, not fourteen tickets.
• NewEvery kind of target and genuine tiered chains. Local disk, SMB, NFS, S3, tape, cloud service and replica — a separate target, cadence and retention per tier, following grandfather-father-son. This makes it possible to express the normal case in a disaster at all: „disk green, tape dead for six weeks". The rule 3-2-1-1-0 is computed rather than ticked and names the reason for every digit not met. For tapes there are media pools, barcodes, write cycles measured against the manufacturer's limit, and off-siting and retrieval with a receipt.
• NewRestore tests as a history of their own. A run with the result „successful" proves that the backup product reported success — not that the data comes back. Type, scope, measured recovery time and measured data loss are recorded; an overdue test colours the plan's traffic light, even if every run was green. If a recovery plan is recorded on the plan, the test appears there as well, and a gap against the commitment moves into information security as a risk.
• NewResults arrive by three routes. By hand (works on day one and for every product, including the tape cartridge from the safe), from the report mail of the backup product through rules per job — unmatched mails land visibly in an inbox rather than in nothing — and through a batch interface for your own scripts. A connector for Veeam is included. In addition: an overview with a traffic light per plan, a capacity forecast per target („full in about 40 days"), a backup policy as a manual chapter generated from the maintained data, and a portal view of your own outstanding reviews.
• ImprovedThe device now says who backs it up. A dedicated section on the configuration item answers the question nobody in the system could answer until now: which job backs up this device, on what schedule, over which chain of copies — and when it last ran successfully.
• ImprovedFour entries on the backup system move across automatically during the update. Retention, RPO, schedule and the last test restore are taken into the new module and kept there from now on — duplicate upkeep disappears. The previous free-text schedule is retained in full; the migrated job is initially dormant, so that nobody finds a list of missed runs on the first morning that they never promised. In the operations manual the previous section is now unambiguously called Backing up the InfraDesk database; beside it stands the new chapter Backup concept for your own backups.

2026.08.10.185910 Aug 2026

• FixedTwo items in the account menu showed a key instead of their label. Instead of My profile and My mailbox (Exchange) a string such as „L9038210eac" stood there. The cause was a mistake in adding the texts to the language files: the four new entries ended up inside another entry instead of beside it. The file remained technically valid, and for an unknown key the display silently falls back to the key's own name — which is why nothing raised an alarm. Only the label was affected; the functions behind it worked correctly throughout. Additionally safeguarded: a check that runs with every change now rejects such entries — in both language versions.
• ImprovedChecklists now live under My work. Until now they had a menu group of their own called Self-organisation — with exactly one item, and under a name that appeared nowhere else in the menu. Beside My work it looked like a second, competing module. A checklist answers the same question as the task board and the task list: what has to be worked through. The group has therefore been dissolved.
• ImprovedThe administration area of the Exchange integration is named like the menu. It sat under Self-organisation — the internal term. It is now found under Settings → My work → Exchange. An administration area with a name that does not exist in the menu is not found: people search for what is on the screen. The manual now names the same route.

2026.08.10.172510 Aug 2026

• FixedThree announced functions were unreachable in the workstation client. In all three cases the function itself was fully present — what was missing was the route to it. The manual described steps at the end of which no screen stood. That is the most unpleasant kind of gap: It looks like a missing function, cannot be worked around, and anyone following the instructions looks for the mistake on their own side. The details are in the following points.
• NewThe Exchange connection can now be set up — under Administration → Self-organisation → Exchange. Exactly this path stood in the manual, and exactly there was nothing until now. Entered are the EWS address or autodiscover domain, server version, sign-in method and the certificate policy; a Connection test names the failed stage in plain words instead of merely „failed“. The operating state stands next to it: last success, last error, back-off period — and a connection shut down after a series of errors can be released again there. Not to be confused with „My mailbox“: That remains every user's personal release. Setting up the connection is a matter for administrators, releasing one's own mailbox a personal decision — two different permissions, therefore two places.
• ImprovedFields that have no effect in the chosen mode are not offered in the first place. No password field next to sign-in through the service identity — there is no password there — and no forwarding of the user sign-in with stored credentials, because there is then no user ticket that could be passed on. The installation would have rejected both anyway; a field whose input is discarded on saving is a trap.
• NewA device model can be assigned to a device — with search and trial run. The catalogue brings over 4,000 models with it, but in the device „model“ was a text field without search, and the assignment could not be triggered at all. The catalogue therefore remained a list, although one row turns into forty-eight maintained ports. Now searching and assigning happen in the device; a trial run shows beforehand which ports would be created, which existing ones remain untouched and which references of the template point nowhere. Existing ports are never overwritten — plugged connections and assigned IP addresses hang on them.
• ImprovedManufacturer and model are locked as soon as a catalogue model is assigned. They then come from the catalogue. Until now they could still be edited while the installation discarded the input — the display therefore drifted silently away from the catalogue. Anyone wanting to set the fields freely releases the model assignment; the ports already created remain in place.
• ImprovedThe account icon is the central place for everything personal. Notifications, one's own mailbox, password and second factor lay scattered among administration pages; anyone looking for „my account“ found a tenant administration. The menu behind the account icon now brings everything together and names in its header whose whose settings are below it — the most important line at shared workstations. Along with it a new page My profile with account data, account type and the sign-in; in the settings the section My account is now right at the top.
• Fixed„Change password“ was offered even to accounts that have no password with us at all. Anyone signing in through the corporate directory (AD/LDAP) or an identity provider has no password entry here — the dialogue therefore inevitably ended with „The current password is incorrect“, which looks like a typing error and is none. The item now appears only for local accounts. That is deliberate and not a limitation: Larger organisations handle password changes through their own self-service portal or a password policy solution, and precisely those systems should remain responsible for it. In „My profile“ the reason is stated in plain words instead of a greyed-out button.

Week 32 · 3–9 August 202615 versions

2026.08.09.191009.08.2026

• NewData centre documentation: what kind of device this is and what is plugged into it. Until now the CMDB answered what you operate and what it depends on. What stands in the cabinet, which ports a device has and where a cable leads was nowhere — each of these questions began with a walk to the server room. New are a Device model catalogue, Device components of every kind, patch panels with pass-through, which Path tracing and the reservation of height ranges. The details are in the following points; a chapter of its own in the manual describes them coherently.
• NewOver 4,000 device models are already entered — without internet access. The catalogue comes filled: 4,289 models from 35 manufacturers including height, depth, airflow, weight and port complement, plus 140 rack models. That is deliberate and not convenience: server environments in municipalities and public authorities frequently have no outbound internet access, and a catalogue that would first have to be downloaded would remain permanently empty there. It is updated through the usual product update. Please note: A product update touches exclusively the delivered entries — whatever you import or create yourself remains untouched, even with an identical name.
• NewThe model brings the ports with it — nobody types 48 of them by hand. A device model carries its complement, and a device inherits it on assignment. Only that makes port management something maintained day to day instead of filled once and forgotten afterwards. A dry run shows beforehand what would be created. Existing ports are never overwritten — plugged connections and assigned IP addresses hang on them; they remain in place and are named in the report. Series can be created without a model too: a naming pattern and „1-48“ are enough.
• NewPatch panels and the question of where the cable ends. A patch panel is not a device with interfaces but a run: plugged at the front, permanently installed at the back. It is now a class of its own, and for every port the path to the far end traced — across any number of panels, with every stop along the way. If the trace does not end cleanly, the reason is stated in plain words: dead end, ring cabling or a split. Nothing is guessed: With a multiply occupied rear side whose fibre does not follow from the course, the display says so — a guessed fibre leads the technician to the wrong port and looks convincing while doing it. Likewise a scheduled partial section does not count as plugged.
• NewReserving positions in the rack before the device arrives. A free position that has already been allocated otherwise leads reliably to the same sequence: two people look into the same occupancy table, both find U 20–23 free, both plan for it — and it comes to light on delivery day. Height ranges can now be reserved with a purpose and a deadline. Occupied and reserved are reported separately and counted separately: With an installation you move aside, with a reservation you call whoever entered it. An overdue reservation no longer blocks but stays visible — otherwise after two years the occupancy plan is cluttered with plans that never materialised.
• ImprovedFull-depth devices now occupy the space behind them as well. Until now the rack occupancy checked the front and rear sides separately. Behind a full-depth server a second device could therefore be entered at the same height — physically impossible, inconspicuous in the table, and it only comes to light once the device has already been ordered. Please note: For installations recorded before this version the depth is not known; they continue to count as not full-depth. Nothing changes in your existing racks as a result — only an assigned device model brings the information with it.
• NewTaking over an inventory from NetBox. Anyone who keeps their physical documentation in NetBox today can take over locations, rooms, racks, device types, devices including their mounting position, components and cabling. This is explicitly a one-off migration and not a permanent synchronisation: There is no schedule and no trigger, the run is started by hand, and after the takeover the connection is switched off and the access key deleted. A permanent reconciliation would make the foreign source a second truth alongside your CMDB. Within the migration project the run is repeatable as often as you like — several attempts are needed until the mapping fits —, every write has a trial run before it, and whatever cannot be mapped is reported and not guessed: a wrongly guessed item type is more expensive in a CMDB than a missing one.
• FixedAfter a successful update the client update service reported a failure. After the file exchange the service checks whether the expected version really is on the workstation now. This comparison ran over the character string — and the two sides wrote the version differently („2026.8.9“ against „2026.08.09“). The exchange succeeded every time, and the service reported an error nonetheless. The comparison is now made numeric. Please note: The correction sits in the service on the workstation — a machine with the old version reports the failure once more while it lifts itself to this version. After that it is settled. The decision as to whether whether an update happens was never affected.
• FixedInstallers carried a shortened version number in the file name. Off 2026.08.09.1353 was 2026.8.9.1353 — a different name for the same state, which meant the comparison with the published version could no longer be carried out. All three packages now carry the number in the same notation as the installation's version.

2026.08.09.135309.08.2026

• FixedNine functions had been announced but were reachable only through the interface. A review of all 103 interface addresses against the user interfaces showed that not a single route led to nine of them in the client or the portal. Affected were, among others, the administration of the identity provider, the release of one's own mailbox and the team boards — all points already named in the previous version of this list. They did work, but only for someone operating the interface directly. That was our announcement, and it reached too far. All nine now have a user interface; the details are in the following points.
• NewSetting up identity providers — with a connection check before it becomes serious. Under Settings → Identity providers OIDC providers can be created and changed: authority, client identifier, e-mail domains and the mapping of the provider's groups to roles. A Probe retrieves the provider's details and shows what it reports about itself. It warns instead of reporting green if a sign-out address is missing or the prescribed security procedure is not offered. The switch „Sign in only through this provider“ warns when switching over against locking yourself out, and names the emergency route. Please note: The provider's secret is never displayed — an empty field means „unchanged“ and deletes nothing.
• NewThe group mapping works itself out before it takes effect. When mapping provider groups to roles, a sample account can be typed in; the preview immediately shows which roles would result — without saving. If a mapping points to a role that no longer exists, the group counts as not assigned: the preview therefore promises no role that is never granted.
• NewReleasing your own mailbox — in the client and in the portal. The release for the Exchange calendar is now where it belongs: under My account in the portal and under My mailbox in the workstation client. The notice states explicitly what the release means — that the service account could technically open every mailbox in the organisation and that the limitation in the product lies. Revoked releases remain in the list: They are the evidence of the period during which access was permitted.
• NewTeam boards are visible — and deliberately read-only. In the task board you can switch at the top between your own board and the team boards. Other people's cards not be moved; dragging is blocked and a bar says why. The reason: every person changes only their own tasks — an action that is offered and subsequently refused is taken for an error and not for a rule. Post holders without a user account are named explicitly, instead of being left out: an empty team board would otherwise look like an error where in truth an account is missing.
• NewAutomation rules for new tickets. Under Tickets → Tools rules of the form condition → action can be maintained: subject, ticket kind, category or reporter domain set priority, group, category or a watcher. The list shows condition and action written out; twenty rule names without this information would be useless. A rule without effect is named as such — especially one that only stops the chain: it changes nothing but conceals all subsequent ones. Not to be confused with the routing rules, which continue to stand alongside and map category to group and nothing else.
• NewAccess review: confirming who holds which role, recurrently. A run collects all role assignments at the start; each is confirmed or withdrawn individually, with a justification. People and roles appear with plain names — anyone asked to confirm an identifier decides blindly. Completion is only possible once no position is open any more: a completed run with open positions would be evidence that proves nothing. This is the control that ISO 27001 requires under A.5.18.
• FixedThe access review collected nothing — every run stayed empty. While building the user interface it emerged that starting a run only changed its state without collecting the role assignments to be reviewed. On a demonstration installation this went unnoticed because its sample data brought the entries with it — in live operation the function was therefore without effect. Also fixed: The individual decisions did not check which tenant the run belonged to. Anyone who knew a foreign identifier could read foreign decisions and take part in them. Both are now safeguarded by checks that run along with every change.
• NewData subject rights and the processor contract register in data protection. There is now a page for the response under Art. 15: enter the e-mail address and the accounts, contacts and tickets of that person appear — exportable as a machine-readable file, as Art. 20 requires. The erasure anonymises instead of deleting: tickets and evidence remain evaluable, the person is no longer identifiable within them. It is only possible after the response has been produced — whoever has seen beforehand what hangs on it decides differently. Alongside it the Processor contract register under Art. 28: the tile „processor contract expired“ previously counted an inventory that could not be maintained anywhere.
• NewIssuing cross-tenant grants. For holding structures and service contracts, a person can be given time-limited access to a foreign tenant. The permissions come as a selection from the catalogue, not from a text field. A grant without an end date is marked explicitly in the list as „never expires“ — precisely this kind of access otherwise stays open because everyone assumes it is settled. The default remains from.
• ImprovedYour own tasks can be edited in full — in the portal too. All columns were visible in the task list, but none could be changed: title, description, due date, urgency and links can now be maintained through a form, and deleting asks first — in the workstation client as in the portal. The portal did not have the list at all until now; it knew only the board. In addition the lead time of the reminders can now be set instead of merely taking effect.
• ImprovedThe manual has become a coherent body of documentation. 36 individual documents became a suite with cross-references, a keyword index and a table of contents per document; along with seven flow diagrams. The occasion was a finding: The overview map named 27 of 36 documents — ten existing ones could not be found through anything, and because each was faultless in itself, something like that never comes to light. Also fixed: the jump marks in the PDF outputs led nowhere.

2026.08.08.205008.08.2026

• NewSign-in through your organisation's identity provider — Entra ID, Keycloak and others. Anyone already operating a corporate sign-in can now use it for Ordivis: the workstation client opens the Systembrowser, the sign-in runs there, and the client receives only the result. Accounts are created at the first sign-in; roles can be assigned from the provider's details. A nightly reconciliation deactivates accounts that no longer exist in the directory. Please note: The client holds no secret of the provider — the exchange runs through the interface. The existing sign-in with user name and password remains possible unchanged; the switch is a setting per tenant, not a requirement.
• NewSelf-organisation: task board, task list and a personal calendar. Between what a ticket system knows and what actually lies on the desk there is a gap — the call without a ticket, the point from a meeting, the „follow up next week“. There is now a place for that inside the platform: a kanban board with freely nameable columns, the same tasks as a filterable and exportable list, and „My day“ as a selection of what is due today. Tasks can be link with tickets, items and changes and created directly from their lists. This is explicitly not a second ticket system: no SLA, no reporter, no escalation. And: Only the person themselves sees their own tasks — administrators do not either.
• NewHanding over tasks — by acceptance, not by assignment. A task can be offer. It only changes owner upon acceptance; if it is declined, it comes back with a justification. Simply placing a task on someone else's board would be assigning work without consent — that is what the ticket system is for. In addition there are Team boards whose members come from a support group or or an organisational unit; they show title, column and due date, no notes.
• NewSubscribing to the personal calendar in your own calendar program. Follow-ups, due dates, appointments, maintenance and change windows are in one place; a click leads into the ticket. In addition a personal address can be generated and entered into Outlook, Thunderbird, Apple Calendar or Android. Please note: The address is a secret — it replaces the sign-in, is shown once and can afterwards only be revoked, not recovered. The retrieval is read-only throughout. Reminders for your own deadlines can be set per user with a lead time.
• NewExchange connection for your own calendar (on-premises). If an in-house Exchange is connected, your own appointments appear in the calendar; on request due dates are also written to to Exchange. Two points that matter for the rollout: the default is the mode „Free/busy only“ — occupied periods without subject, location and participants; and every user releases their mailbox themselves; without this release no mailbox is opened, not even an administrator's. Please note: Anyone switching on writing to Exchange and and also using the subscription sees every appointment twice — both routes are deliberate, you decide on one. A two-way reconciliation does not take place.
• NewTask board and calendar in the self-service portal too. Both are now available in the browser as well. Deliberately without drag and drop: Moving happens through a menu on each card. That can be operated with the keyboard, on a phone and with a screen reader — for a portal whose purpose is occasional access, the better trade.
• NewRack occupancy with real installations instead of free text. Until now a device's position stood as free text on the item („U 12“, „12-14“, „right at the top“). Nothing could be checked against that: two devices could claim the same height, and a 4U server fitted arithmetically into a full rack. An installation is now a datum of its own with a height range and a side, with collision and capacity checking; the occupied rack units are calculated instead of maintained. The existing inventory can be taken over — whatever cannot be read unambiguously is not guessed but shown as a list for follow-up work.
• ImprovedObject-level rules previously took effect only on creation, not on reading. Anyone who had stored rules at object level in the permission administration („cost centre 4711 only“, for example) was entitled to assume they applied on View as well. They did not: the check ran before loading and saw only what the request itself brought with it — enough on creation, not enough on reading. Item details and the dependency graph now check every object individually. Without a stored rule nothing changes, neither in the result nor in the speed. Please note: Concealed objects are conceals and not omitted — a node in the graph and an installation in the rack keep their place and lose only their content. A gap would be more dangerous than a grey area: a technician miscounts past it.
• ImprovedSelf-organisation in three languages, with a manual chapter and F1 help. The three new views were initially built firmly in German; they are now translatable throughout and stored in German, English and French — including the dialogues. Along with them a chapter of the manual and the context-sensitive help on F1. Please note: The application's language selection remains German and English — the comprehensive extension to further languages is a step of its own.
• FixedFurther lists in the workstation client showed entries twice. After the finding in the organisational structure, the same construction was searched for and found in 16 further views as well: the list was cleared before the retrieval and filled afterwards; if two retrievals overtook one another, both entered their result. They now swap their lists only after the retrieval. Your data was never affected — it was exclusively the display, and a refresh already cleared them beforehand.

2026.08.08.130208.08.2026

• NewEvery application logs by the same rules — and one click packages the state of the installation for support. Until now every application wrote its log by its own rules, and some wrote none at all. Under „Logs“ the super admin now sets centrally how verbosely logging happens, how long the files are retained and when rotation occurs; the minimum level can be set separately for 28 areas and takes effect without a service restart. The applications that were missing until now have been added: wallboard, client update service, updater, update indicator and update agent. The Support package collects logs and key data into an encrypted archive. Please note: The password does not belong in the same message as the archive.
• FixedThe interface's logging settings never had any effect. This is the more serious part of the finding above: the delivered configuration of the interface contained a second log sink pointing at the same file as the built-in one. The second held the file exclusively, the built-in one failed silently — without an error message. Its hard-coded values therefore applied, and everything a super admin set was stored and remained without consequence. Demonstrated on the running installation: not a single line in the pattern of the built-in sink, 821 in the pattern of the configuration. It went unnoticed in the portal — there was no second sink there. An automatic check forbids such sinks in configuration files from now on.
• FixedAgents could not get at their own tickets in the portal. The overview tile showed „7 open tickets“, the list one click further reported „No tickets found“. These were not two different sets — both routes narrow the same way; the list simply received no clearance. The permission intended for it was granted only to the role „End user“: 10 of 13 supplied roles were affected, among them service desk L1/L2, CMDB and network administrator, change manager, CIO, data protection and information security officers. Every role now carries it; existing installations catch up on update. Nothing is opened up by this — the query continues to narrow to one's own tickets. Still open: A self-created role without this permission still runs into the same refusal; the permission can be added in the role administration.
• FixedA refused session renewal in the portal set off an avalanche. If a session was ended, all waiting calls sent their own long-since hopeless renewal after it: in the log 112 attempts in one second. Only that made the interface's overload protection take hold — the refusals were the consequence, not the cause. The session now remembers the refusal itself, and whoever arrives afterwards aborts. Also fixed: such a refusal could not be attributed to anything afterwards — it appeared in no log. Path, rule, origin and waiting time are now recorded.
• FixedPosts appeared twice in the organisational structure. After creating a post or filling a management post, every post of the unit stood twice in the list — with two posts therefore four entries. In this case the view reloaded twice simultaneously and entered both results. Particularly misleading: when filling a post without without the management attribute it did not occur, so the duplicates apparently disappeared again by themselves. Your data was never affected — it was exclusively the display; a refresh already cleared them beforehand.
• FixedThe web portal showed every post without a name and without exception as „vacant“. Under „Organisation“ the column „Post“ stayed empty, and all posts carried the tag „Vacant“ — the filled ones too. The portal expected field names that the interface does not supply; an unknown field name does not lead to an error message but to an empty value. The page therefore looked impeccable: a filled table with empty cells. Now the post designation, holder and deputy stand there; shared appointments with their share, and „vacant“ means vacant again.
• Fixed„SLA breached“ meant something different in four places. The tile stood permanently at 0 — it counted from a table into which nothing was ever written. The ticket list next to it calculated for itself and coloured everything red whose deadline lay in the past: on the data of a test installation that was 154 tickets, breached were 61. Among the false alarms: 39 tickets closed within their deadline and 54 in waiting states in which the SLA clock is paused. Tile, list and red marking now mean the same thing — the determination of the evaluation run, against which SLA attainment is also measured. Clicking the tile filters on the server, so it no longer shows only the hits of the first page.
• ImprovedThe workstation client now says why it signed you out — and a disruption no longer signs you out. If the session was ended (only one sign-in per account is ever valid), the client cleared it away silently: the window stayed open, and every page failed individually with its own message. It now returns to the sign-in screen and names the same reason as the portal. More important still: Previously ended every unsuccessful response ended the sign-in — including a brief overload or a restart of the interface. Only a rejected renewal token still leads to a sign-out.
• ImprovedThe business allocation plan explains how a responsibility comes about. The tab „Responsibilities“ showed three buttons and a list, but nowhere the rule behind them. It now names the order of precedence for the lead, every button says what it does, the state Proposal or confirmed can be recognised at a glance, and a status line names the next step. Nothing changes in the behaviour — the rule applied in exactly the same way before and stood in the manual too; it simply was not visible in the application.
• ImprovedThe Android app and the manual carry the Ordivis brand. The app's start screen has been reworked: scanning and reporting move down within thumb reach, the tiles stand in one row, the connection dot sits in the header — together around a third less height. In addition 46 illustrations in the manual now carry the current brand.

2026.08.07.210407.08.2026

• FixedEnd users never saw their own tickets in the self-service portal. Anyone signing in as an end user got an empty page under „My tickets“ — regardless of how many tickets they had reported. The cause did not lie in the display but one level deeper: the interface demanded the full read permission across the board for the ticket list, while the role „End user“ deliberately carries only the permission for their own own tickets. The permission was therefore granted, but no endpoint accepted it. Three calls that the portal makes for every end user were affected: the list, the detail view of a ticket and the service catalogue. All three are now bound to the rule that was stored anyway. Who sees what remains unchanged: The complete ticket list continues to require the full read permission, and someone else's ticket remains invisible to end users.
• Improved„Why was I signed out?“ — the portal now says so. Only one sign-in per account is ever valid; a new sign-in ends the previous one. That is deliberate and stays that way. The portal, however, reported this case with the sentence „Your session has expired“ — and that was practically never true: it would only have expired after eight hours, it was ended by the second sign-in. Both cases are now distinguishable and are named differently as well. Please note: The short-lived access token remains valid for up to 15 minutes after revocation — the sign-out on the first device is therefore not noticed immediately but only at the next renewal.
• FixedGenerated documents carried the old product name as the company name. The header of every generated PDF carries the tenant's company name. If none is stored, a fallback value steps in — and after the renaming that still read „InfraDesk“ in several places. Affected were reports and printouts of tenants without their own branding as well as the legal block of the e-mail signature; that is, precisely the documents that leave the organisation. The product name now stands in exactly one place in the code. Not changed, because deliberate: The attribution „Created with Ordivis Platform“ in the footer of every report is retained.
• FixedThe authenticator app showed the old product name. When setting up the second factor the account name is taken over into the app; there it still read „InfraDesk“. Newly set up devices now carry „Ordivis Platform“. Devices already set up keep their existing label — it is stored in the device and changes only on re-setup. The sign-in itself is unaffected, existing codes remain valid.
• FixedRaw timestamps in the crisis team journal and on the process tasks. In both places the technical date format appeared instead of the customary notation. Also fixed: a raw date in the ISO 27001 view, a wrongly labelled column in the distributed discovery and a window title that still showed the old product name before sign-in.
• FixedThree small things in the portal that got in the way day to day. The badge with the number of unread notifications was cut off by the header bar. A heading contained a double asterisk from the markup. And the empty password field looked, because of its placeholder, as though something had already been entered.
• ImprovedRetention for the toolbox installers too; SLA tile and organisational data in the demo tenant. Old toolbox installation packages were never cleaned up because they follow a version sequence of their own — they are now limited by version like the other packages. In the demo tenant not a single ticket had a deadline, which is why the SLA tile stayed permanently empty; likewise the organisational units lacked their type. Both have been added — the tile was never broken, it was missing the data.

2026.08.07.054907.08.2026

• NewAn area for your own work: task board, task list and calendar. Until now Ordivis could tell you very precisely which Tickets are open — but there was no place in which to plan your own day. Anyone wanting to note something that is not a ticket — „call back company X“, „inspect the server room“ — did that alongside: on paper, in the mail program, somewhere. Under „My work“ you now find three new views. The Task board shows your tasks in five columns that you change by dragging; the column „Waiting“ separates „I am stuck“ from „I have not started yet“. The Task list shows the same tasks as a sortable, exportable table. Every task can be linked to a ticket, a configuration item, a change request, a problem or a knowledge article. Your tasks are visible exclusively to you — an administrator does not see them either.
• NewThe calendar brings together what was previously spread across six pages. Follow-ups, ticket due dates, deferred tickets, agreed appointments, maintenance windows, planned changes, workflow deadlines and your own tasks now stand in one view — as month, week, day or agenda. One click leads to the ticket. The calendar stores nothing of its own: It shows the appointments that hang on your tickets anyway. Anyone wanting to change a date changes it on the ticket — so there continues to be only one truth about when something is due. You see only what you are allowed to see anyway: without read permission for change requests, their windows do not appear in the calendar either.
• ImprovedFirst stage of development — we say openly what is still missing. Cards can be dragged between the columns, not yet reordered within a column. The task list can sort and export but cannot yet filter or group per column. The calendar does not yet know your working hours and public holidays and therefore does not highlight weekends. These three points will follow; the area is usable without them too.

2026.08.05.210205.08.2026

• FixedThe ticket list never showed the reporter's company. In the column „Reporter“ the company is provided for next to the name — it stayed empty because the server did not send this information at all. It therefore looked as though not a single reporter had a company stored. The company comes from the contact directory and therefore concerns external reporters; your own staff have none, and that is deliberate — it is precisely the company that distinguishes the external reporter from your own personnel.
• ImprovedThe update manifest carried a component that never contained anything. Since 07/31/2026 the description of every update contained an entry „worker“ with zero files. The associated service had long since been replaced — the application does the background work itself — but the empty entry remained. Nothing changes for you in the update; the description now names only what is actually delivered. An empty component cannot be distinguished in the description from a working one: it looks like coverage and delivers nothing. In future the build reports that loudly instead of carrying it along silently.

2026.08.05.191405.08.2026

• FixedA changed priority was not saved. This is the reason for marking it a mandatory update: an input was lost, and it did not look like it. The display jumped to the new value immediately, and the ticket history even showed the entry „Priority: medium → high“ — only nothing went to the server. At the next opening the old priority stood there again while the history continued to assert the change. The sidebar saves every field on leaving it; precisely with the priority this trigger was missing — it was the only field without one. Anyone who happened to touch another field after changing it triggered the save along with it: that is why the error occurred irregularly and looked like a display problem of the list. An automatic check now ensures that no field of the sidebar is left without saving.
• NewA ticket can be split. Reporters often write several topics into one e-mail. In substance these are separate tickets — different responsibility, different priority, their own deadline. As long as they are stuck in one ticket, the ticket only counts as done once the slowest topic is finished, and the deadline of the quick one is measured against the run time of the slow one. Through „Actions → Split ticket“ a topic now becomes a ticket of its own: with its own deadline, linked to the original and recorded in both tickets. The original stays open — whether anything remains to be done there is decided by the agent, not by the system. The previous history stays with the original, where the mail with both topics is; the link leads there from both sides.
• NewKeywords in the subject can set the category. A large part of the reports carries the topic in the subject already — categorising by hand nevertheless costs time on every ticket and turns out differently each time. A rule such as „subject contains Printers“ now sets the main and subcategory itself. The benefit lies not in the tidiness but in the responsibility: The assignment to the support group sits on the category, and it is evaluated again after the setting. With an e-mail that is the difference between a ticket that waits and one that is worked on. If the reporter has specified a category themselves, that one stays.
• FixedThe column for major incidents in the ticket list always stayed empty. The star indicates that a ticket has been declared a major incident. The column existed together with its legend, but the list did not receive this information from the server at all — so it could never light up, no matter how many tickets were declared.
• ImprovedNo more releases without completely green checks. Both test series now run before the first build step; a failed check aborts the publication. The occasion was a finding of our own: one of the series was red for three days without it being noticed. The failures themselves were harmless — but a real regression could just as easily have sat in the same place.

2026.08.05.152605.08.2026

• FixedA repeat installer run could still lose your HTTPS configuration. The safeguard from the previous version was faulty — in three respects, any one of which would have sufficed. First, the certificate path was indeed pre-filled from the existing configuration, but read incorrectly: it came back with doubled slashes, so the service would have looked for a file that does not exist under that name — and would then have fallen back silently to unencrypted. Second, the safeguard cancelled itself out: it determined whether the administrator had entered something in the wizard, but checked this only after it had already inserted the existing value — and therefore wrongly took it for an input. The mode consequently stayed at „reverse proxy“ and the installation continued to run unencrypted. Third, an installation script lacked a file marker without which Windows may not be able to execute it at all.
• ImprovedThe installer now says when the certificate is not where it should be. If the entered path points nowhere, that previously appeared only in the application's log — and then only after the start, when nobody is looking any more. It now appears in the installation log while the setup is still running.
• ImprovedThe safeguard is now checked automatically. Four sequences are played through on every change: clicking through the wizard (nothing may be lost), deliberate re-entry (that applies), fresh installation, and the deliberate switch to reverse proxy. What is checked is the code actually delivered — a copy of it behaved differently and would have let the error through.

2026.08.05.142105.08.2026

• FixedOutgoing e-mails did not reach the recipient as soon as more than one tenant was configured. This is the reason for marking it a mandatory update. Replies to reporters, password reset mails, appointment invitations — everything stayed in the queue, although the mailbox was configured, active and reachable. The send run works per tenant, but pulled the messages all tenants out of the queue; the search for the matching sender mailbox, by contrast, stayed with the tenant currently being processed. For a message belonging to another tenant it therefore found no mailbox and fell back on an emergency route that does not exist in normal operation at all. With four tenants three out of four runs each burned one delivery attempt — after six attempts the mail counted as finally failed, and the one correct run never got its turn because of the growing waiting time. The more tenants, the more certain the failure; with a single one it does not show up at all.
• FixedA repeat installer run took away your HTTPS configuration. The wizard always started at „reverse proxy“ with empty certificate fields and wrote exactly that into the configuration — domain, certificate path and password were gone afterwards, the installation ran unencrypted again. Particularly unpleasant: a mandatory update is applied through exactly this route. The wizard now shows your existing configuration and takes it over unchanged if you change nothing. The password is retained if you leave the field empty — it is stored encrypted and is not readable by the installer at all.
• NewCertificates can be pasted as text — no more converting by hand. Certificate, private key and issuer chain are simply copied in from the files of your certification authority. Until now you first had to build a PFX file from them and type in its path; a mistake in doing so only came to light at service start, where the application fell back silently to HTTP and the reason stood only in the log. It is now reported immediately whether certificate and key match, until when it is valid and for which names. You do not have to assign a password for this: the private key is stored protected and is nowhere held in clear text.
• NewThe self-service portal can do HTTPS with the same certificate as the API. Until now portal TLS could not be configured permanently — the only possibility was a file that every update overwrites. Portal and API now share mode, domain and certificate and merely have ports of their own. Port 443 is recommended: the portal is then reachable without specifying a port — the address you pass on to your workforce. Existing installations remain unchanged; you switch portal HTTPS on deliberately.
• FixedLinks in e-mails pointed to „localhost“. Password reset, accepting or declining an appointment, giving a rating — these addresses worked exclusively on the server itself, so never for the recipient. They could not be corrected permanently: the value stood in a file that every update replaces. As soon as HTTPS is configured with a domain, the addresses are now derived from it.
• FixedThe tile „Major incidents“ showed tickets that were not any. What was counted was the highest escalation level — and the deadline monitoring raises that by itself without anyone having declared anything. An earlier error additionally raised it every minute, so that a briefly overdue ticket immediately stood at the very top; it never went back down. Conversely a really declared major incident was not counted at all, and „revoke“ revoked nothing. What is counted now is the declaration itself. A side effect that weighed more heavily than the display: Tickets at the highest level are skipped by the escalation — they would never have triggered a notification again. This legacy is cleaned up during the update.
• ImprovedA switched-off mailbox can no longer be overlooked. The state stood as ordinary text next to „OK“; that no e-mails are being fetched at all as a result stood nowhere. It is now highlighted in colour and names the consequence. In addition the log reports once if no receiving mailbox is active at all.
• Fixed„Test connection“ delayed the inbox by a full retrieval interval. The test set the same marker as a real retrieval — the next collection was thereby postponed by up to 15 minutes, and during setup you test several times. Precisely where you are waiting for the first mail, the delay was greatest.
• FixedThe update window could not be closed during the follow-up phase. It went on waiting for the application afterwards — and this waiting time could run into nothing and last around five minutes although the update had long since finished. Closing is now possible as soon as the update is through.
• SecurityThe program directory is hardened during installation. If the installation lies on a data drive, the folder inherited its permissive permissions: every signed-in user could exchange program files — which then ran with system rights. Only the system and administrators now have write access.

2026.08.05.081105.08.2026

• FixedAn update could roll itself back although the server was perfectly healthy. The final health check gave every request five seconds. The endpoint /health however combines all check points and answers only once the slowest is through — if, for example, no mail server is reachable, this point alone costs around three seconds. On a machine with a 5.1 second response time every single request therefore ran 100 milliseconds too long: not an occasional lapse but a guaranteed failure. The update was applied, the check failed, everything was rolled back — anew on every attempt. The time limit is now 30 seconds, likewise in the installer path (there it was as little as four).
• FixedThe waiting time was governed by the error pattern instead of by the matter at hand. Waiting happened in 20 attempts with a two-second pause each. If a connection was refused immediately, it was over after 41 seconds; if it ran into the time limit, only after 141. But what is being waited for is something specific: the database migration at the first start after an update, and that can take minutes. A fixed period of five minutes now applies — the same one the installer has always allowed. Every 30 seconds the log reports that waiting is still going on, and for how much longer.
• ImprovedIf the health check fails, the reason appears in the log. Until now it merely said „health check failed“ there — without saying why. Whether the port was closed, a certificate did not match or the time simply was not enough could not be distinguished.
• FixedThe password of the certificate file is now protected after a fresh installation too. The previous version encrypted it when saving in the settings — but the installer stores it in clear text, so it came back after every setup. The service now catches up on this itself at start and records it in the log. It cannot be solved in the installer itself: it runs under your administrator account, the service as the system account — a value encrypted there would later be unreadable for the service.

2026.08.05.065605.08.2026

• FixedInstallations with their own certificate could no longer update themselves. The update ran through completely, failed at the final health check and was rolled back — anew on every attempt. The application was at no point unhealthy: the check queried http://localhost:5000 in a hard-wired way, although in pure HTTPS operation no HTTP endpoint is bound at all. It therefore ran against a closed port. Affected were, of all people, those who had configured TLS properly. And the reason stood nowhere — the log reported only „health check failed“ without saying why. The address now follows the actual configuration: HTTP as long as it is bound, otherwise HTTPS on the configured port. The certificate check is exclusively skipped for localhost — the certificate is issued for your real host name and does not cover „localhost“. The reason for the failure now ends up in the log.
• FixedThe password of your certificate file stood in clear text in the configuration. In hosting.json below ProgramData the PFX password was openly readable. Whoever has it can extract the private key from the PFX file and impersonate your server. What is remarkable about it is not the missing procedure but that one existed: the backup password in the same store had always been protected — here only the same step was missing. The password is now encrypted with Windows DPAPI, readable only by the service account; a clear-text value that is still present is caught up automatically at the next save in the settings. Existing installations continue to run unchanged — a clear-text value remains readable so that this update switches off nobody's TLS.
• ImprovedThe port of the health check comes from the hosting settings instead of from a fixed number. In this place there was previously a condition whose two branches delivered the same value — it looked as though it read the configuration, and never did.

2026.08.04.203704.08.2026

• FixedYour backups were stored unencrypted although a password was on file. This is the reason why this version is marked as a mandatory update. The backup script looked for the settings file in a folder into which it has no longer been written since the renaming. It therefore never found it, evaluated that as „no password on file“ and stored the backup in clear text — with a note in the log that nobody reads in everyday work. A missing path and a missing password could not be distinguished, and in case of doubt the script decided in favour of the insecure variant. Every backup created since 3 August is affected: they are complete database dumps. Check your backup directory for files with the extension .dump, create a new backup after the update and remove the old ones. The path that was checked now appears in the log.
• FixedAn upgrade could switch the application over to an empty database. The installer entered the database name in a hard-wired way. After the renaming this name no longer matched existing installations — the upgrade created a new, empty database alongside and pointed the connection there. In doing so nothing deleted, the previous database remained in place untouched; it merely looked as though all the data were gone. The name is now taken over from the existing installation, likewise in the setup scripts run by hand.
• FixedOld backups were never deleted. The cleanup filter looked for the old file name, the new one was written. The configured retention period was therefore without effect and the directory grew without limit.
• FixedRestoring a backup appeared to be missing. The overview recognised only backups with the old file name and therefore stayed empty — and without a row there is no button. The route was fully present the whole time. Backups of of both naming generations are now found and restored, including those from before the renaming.
• FixedAn account that receives its role through a directory group saw an empty program after signing in. The permissions did arrive, but only at the next directory reconciliation — and the navigation, already built, never evaluated them again. It looked like a permissions error but was a display error. The navigation now rebuilds itself automatically as soon as the permissions arrive, and tells you so as well. Anyone saving the group mapping thereby triggers the reconciliation directly instead of waiting up to an hour.
• FixedAn upgrade reset the backup directory to the default path. Anyone who had configured a directory of their own was no longer offered it in the wizard; anyone who clicked through then wrote the daily backups somewhere else — the existing ones had disappeared from the interface. The configured path is now taken over.
• FixedThe tile Portal service in the platform monitoring stood permanently at „Unknown“. It queried a service name that exists on no installation. It would therefore never have turned red even during a real outage — a traffic light that knows only green and grey is none.
• FixedSecurity events were missing from the Windows event log. Two of the three setup routes did not create the event source. If it is missing, the application does not write its security events — and does so silently, because it deliberately does not turn the absence into an error so as not to block any sign-in.
• FixedAfter uninstalling, firewall openings remained in place. Only the rules under the new name were removed; those created earlier remained behind as an inbound opening. In addition a further opening for the same port was added on every installer run. A scheduled task also remained behind until now and afterwards pointed at a script that no longer existed.
• FixedThe hardening instructions in the operations manual were without effect. For the maintenance connection they named a configuration key that the application did not read at all. Anyone who followed them and consequently withdrew the right to create databases from the application role could afterwards no longer create a tenant — with precisely the error message whose remedy the instructions had just withdrawn. Both notations now apply.
• ImprovedManual and help name the actual paths and names again. Tenant databases, the storage location of the update logs and the backup building block in the emergency manual deviated from reality — in an emergency one would have looked for the backups in the wrong place.

2026.08.03.203203.08.2026

• FixedThe client could terminate when switching a saved ticket view. Without a message, without an entry in the log — the window was simply gone. The cause was a feedback loop in the status filter: the tick reported every change back, and the report tipped the value instead of adopting it. If a saved view set the tick, the field followed, reported back, the value flipped, the field followed again — a loop with no exit that terminates the process with a stack overflow. Every route that sets filters automatically was affected: applying a saved view, quick filters from the dashboard, Reset filters. It only became visible if the status filter had been opened once beforehand — before that the ticks do not exist at all, which is why it appeared to occur arbitrarily.
• FixedTwo figures on the ITSM dashboard did not match their label. The ring is called Open tickets by priority, but queried without a status and therefore counted every ticket — including long-closed ones. An installation thus saw „Critical 7“ while not a single open ticket was critical. And the tile Open tickets counted only In progress and Waiting for customer — without New, Waiting for service provider and Blocked. Both now follow the same definition as the wallboard and the evaluation: open is everything that is neither resolved nor closed. Clicking the tile opens the same set that it reports.
• FixedThe button Active Directory on the sign-in screen did nothing. It stood there because the sign-in screen showed a button for every configured provider — including the directory service that the AD connection enters there. But a directory service is not a browser procedure; there is nothing to redirect to. The click nevertheless opened an address, and that answered with HTTP 404. Directory accounts sign in unchanged through the ordinary screen above it. Two further errors on the same route have been fixed along with it: the address for genuine SAML providers was formed incorrectly, and an incorrectly configured sign-in reported Server error instead of saying what is missing.
• NewColour scheme Ordivis — and it is the new default. Turquoise and navy of the corporate design, in a light and a dark variant, in the client, the portal and the app. After installation it is set; anyone who has chosen a scheme themselves keeps it — the default applies only as long as no choice has been made. The previous six schemes are retained in full. Two of the colour pairs originally designed failed the contrast check (3.31:1 and 4.31:1) and were replaced; all text pairs meet WCAG 2.1 AA.
• ImprovedThe brand colour now reaches your recipients as well. Appointment invitations by e-mail, the ticket history in notifications and the appointment form still carried the old blue — they are created on the server and were not covered by the changeover of the interface.
• FixedThe link to the website in the settings still pointed to the old address.
• ImprovedThe manual and the in-app help describe the new scheme. section Appearance and accessibility now names seven schemes and states explicitly how long the default applies.

2026.08.03.184903.08.2026

• NewThe product is now called Ordivis Platform. The name comes from ordo (order) and visus (sight) — it describes what the software achieves: making visible what exists, how it hangs together and who is responsible. The previous name was doubly taken: two other manufacturers carry products in the same field under the same designation. That would have created a need for explanation in tenders at the latest. Nothing changes in your data — database, backups and licence remain untouched.
• ImprovedNew visual identity in the client, the portal, the app and the manuals. Symbol, wordmark and colours follow the new corporate design (navy and teal). In the process two legibility faults came to light that existed already before — the labelling on coloured buttons reached only 3.9:1 and thus lay below the accessibility requirement, a muted text colour on elevated surfaces as little as 3.75:1. Both have been fixed along with it; all pages now meet WCAG 2.1 AA.
• ImprovedNew internet address: ordivis.eu. The website, manuals, download area and the update feed are there. You can reach support and sales at support@ordivis.eu and vertrieb@ordivis.eu.
• NewThe modules now carry names. Ordivis Service (tickets), Config (CMDB), Assets, Network (IPAM), Security (ISMS) and Continuity (BCM). The familiar technical terms remain alongside — the names are added, they replace nothing.
• ImprovedThe manuals and the in-app help have been completely reset. 37 PDF documents, the complete manual with 359 pages, plus 35 help topics in the client and the portal.
• FixedThe autostart entry of the display device left a corpse behind after a renaming. The entry in the Windows autostart is now called Ordivis Wallboard; the old one is removed in the process. Without that there would be two entries there, and the old one pointed at a file that no longer exists.
• ImprovedTechnical identifiers deliberately remain unchanged. Database name, service names, the data folder under ProgramData and the connection string continue to carry the old name. That is deliberate: renaming them would have a benefit of zero and the risk of separating an installation from its own data.
• NewThe changeover happens through the installer, not through the update check. The reason is technical: a delta update exchanges only program files and deliberately leaves your configuration untouched — but the new server address stands precisely there. Through the usual update the installation would therefore continue to query the old address. The installer writes it along.

Week 31 · 27 July – 2 August 202621 versions

2026.08.02.174002.08.2026

• FixedReady-made reports, checklist templates and specialist catalogues were missing from your installation — although they were delivered. The report was „the button + New report does nothing“. The button was fine: it found no report to schedule. The supplied inventories were indeed created at start-up, but written into the wrong database — with the correct tenant identifier, only in the central one instead of yours. Every one of these operations reported success. Affected were the report library, the checklist templates, the threat catalogue, the ISO 27001 catalogue, the basic data protection catalogue, the BSI profiles and the sample organisational structure. With this version they are added at the next start; entries that you deleted yourself do not come back in the process.
• NewWindows sign-in without a password now in the web portal too. Anyone sitting at a domain machine on the internal network is signed in automatically when opening the portal; if that fails, the ordinary sign-in stands ready unchanged. This is not a function that runs by itself: The browser negotiates the sign-in with the Portal rather than with the interface, and this identity cannot be passed on. The portal therefore identifies itself to the interface with a shared secret. For this four settings are needed that apply together — operations manual 3.4. If one is missing, sign-in with a password remains. The list of permitted origin addresses is deliberately pre-filled as empty, and empty explicitly means from the same machine only: the secret alone is not meant to suffice. Tenant selection, second factor and the emergency account continue to go through the ordinary sign-in.
• FixedThe ticket number and subject were missing in the portal — and internal notes looked like replies to the reporter. Three fields carried different names in the portal than in the interface. An unmatched field name leads to no error message: it silently leaves an empty value behind. The ticket number and the subject therefore never arrived — not in the heading either — and in the history an internal note could not be distinguished from a reply to the reporter. This had existed since the first version of the ticket page. The history is now additionally coloured as in the client: reporter, agent, internal note and system entry can be told apart at a glance.
• FixedThe portal's status filter offered a status that does not exist. Anyone choosing „Open“ in the ticket filter got always an empty list — the interface does not know this status. Two further statuses had no German label at all and stood there as an English raw value. In addition the portal's status colours drifted apart separately in three places and matched the client in none of them.
• NewReporting errors — from the portal too. As in the client, a button collects the details that support asks for anyway: version, page, point in time and the session's last error messages. Without a screenshot, and the dialogue says so as well: a browser can only hand over the entire screen, not the window — that would be more than an error report needs.
• ImprovedThe ITSM area now stands directly after CMDB — the same in the client and the portal. Anyone using both interfaces does not have to search twice.
• ImprovedThe change log now says what is mandatory and what the minimum version is. These are two different statements and therefore get different badges together with a legend: a mandatory update takes effect only as long as it is the latest version; the minimum version stays. In addition the list loads faster — the client keeps a copy of its own per server and on opening merely asks whether it is still correct.
• FixedThe CMDB import templates contained sample values that the import itself would have rejected. Five errors across all three templates — invalid yes/no entries, duplicate columns, missing mandatory fields. The most expensive was the silent one: In the column Deputy stood a person's name, although yes or no is expected there. The import did not object to it, read the name as „no“ — and the deputy became a holder. A trial run with zero errors never finds something like that. The import dialogue now also fetches its field list from the server instead of carrying a second list that deviated from it.
• NewLabels with UHF RFID: recording a whole room in one pass. Complementing NFC, which is held up to each device individually. The gain is not recognising a particular device — the sticker does that better — but the reconciliation: what is missing and what stands here that does not belong here. Labels recorded more than once are counted rather than discarded; without that number, zero hits would leave it open whether the room was empty or the reader switched off. Please note: The sequence and the data stream have been tested, not the interplay with a device — we had neither a UHF reader nor an RFID-capable printer available. Test with your hardware before switching over to it.
• ImprovedAndroid app: emergency manuals, checklists and the service catalogue — the essentials without a network too. Approved operations and emergency manuals can be taken beforehand onto the device, and every row says whether it is there — an emergency manual is needed when power or network are gone, that is, precisely when reloading is no longer possible. Assigned checklists are ticked off without a connection; the tick appears immediately, and the row says that it has not been sent yet. Orders can be placed from the service catalogue, with processing time and approval requirement before of the order. The app is still not published and not part of the delivered scope.

2026.08.01.190301.08.2026

• FixedThe Windows sign-in could block the sign-in screen. In the midday version a spinner with „Windows sign-in“ turned after the user name was entered, and it went no further. Two causes: the silent sign-in attempt replaced the form and restored it unconditionally — anyone who had meanwhile moved to the password field got the switch in the middle of their input. And its time limit of 20 seconds ran out fully if no domain controller was reachable, over VPN or from home, for example. This version should therefore be applied if you have version 2026.08.01.1324 in use.
• ImprovedSigning in with the Windows account now happens before the screen, not beside it. To the question „does a user name still have to be requested at all?“ the answer is: no. The attempt takes place on opening, with a start display of its own reading „Signing in as DOMAIN\user“; if it succeeds, the screen is never created. A form that you first see and then have skipped for you is not single sign-on. Switching users remains possible: Signing out leads directly to the screen and bypasses this route — otherwise the Windows account would immediately have signed the just-signed-out person back in. Two limitations from operations were added: when signing in with a local Windows account nothing is negotiated in the first place, and the server can be limited through Identity:WindowsSso:AllowedNetworks to particular networks. If it fails, that is now stated — with an interpreted status code in the diagnostic log instead of wordless nothing.
• FixedRate limits took hold in the wrong place — in everyday use as error 429, in the worst case as a sign-in blocked for everyone. Two related errors in the interface, both present for some time and made visible only by the expansion of the portal. First, the counting was evaluated before the sign-in check: the generous limit for signed-in users was thereby never reached, everyone shared the limit for the unauthenticated — and did so per address, behind a portal or proxy therefore one limit for the entire organisation. Ordinary page requests thus ran into a refusal. Second, the limit for sign-in attempts was a only for the whole installation instead of one per sender: eleven sign-ins in the same minute sufficed, and anyone doing that deliberately locks everyone else out of signing in.
• NewThe web portal has become a full workplace, no longer merely self-service. Until now the rule was: the Windows client can do everything, the portal is the lean view for end users. That no longer applies. Agents work on tickets, problems and changes in the browser including internal notes; specialist users maintain the CMDB, locations, assets, licences, contracts and address administration; ISMS, emergency management, data protection and organisation have their working views; reports can be assembled, users, roles and tenants administered. The navigation carries the same areas in the same order as the client — anyone using both does not have to search twice. Tables now have column selection, filters, grouping, saved views and export to CSV, Excel and PDF in the browser too. What the browser cannot do is stated alongside: Windows sign-in without a password, label printing through the local interface, NFC and inventory taking on the local network hang on the workstation itself; the portal points this out in the relevant place instead of leaving them out silently.
• FixedThe portal signed every user out after 15 minutes — or, worse, showed them empty lists. The access token expired after a quarter of an hour and was never renewed; the associated renewal token did lie in the sign-in cookie but could not be redeemed from there at all. Anyone working in the portal for longer than 15 minutes then saw a page that looked signed in but had no content. The session is now kept server-side and renewed in good time; if it does expire, the portal takes you to the sign-in screen and says why there.
• NewWeather and official severe weather warnings from the German Weather Service. Under Weather show the current situation, a three-day forecast and — right at the top — the official warnings with level, period and the recommended actions in the DWD's own wording. The warnings are the actual purpose: For a municipality they are a trigger in civil protection, not decoration. The entry therefore stands in the menu without a permission check — a storm concerns everyone in the building. Station and place name are set in the portal's configuration; without internet access the display can be switched off there. Only the configured station identifier is transmitted, no user or device data.
• FixedWhite type on light green and light blue status badges was not legible in the dark appearance. The badge's colour changed with the scheme, its labelling stayed white — in dark mode that came to a contrast of 1.7 to 1 instead of the required 4.5. The automatic check nevertheless reported „everything passes“: it checked a colour pair that does not occur in the interface at all. Both have been fixed — the label colour is now generated along with every status colour and checked with it.
• ImprovedBehind a reverse proxy the interface now needs a setting — please enter it once. If a proxy stands in front, the interface until now saw the same address for every request: that of the proxy. Everything that counts per address therefore shared one common count for the whole organisation. Enter the addresses of your proxies under Hosting:TrustedProxies in (operations manual 3.2). The list is deliberately pre-filled as empty — as long as it is empty, forwarded origin details are ignored. If the interface believed them unchecked, any caller could claim an arbitrary origin and circumvent the limits. Without a proxy there is nothing to do.

2026.08.01.132401.08.2026

• NewSigning in with the already signed-in Windows account. Anyone sitting at a domain machine will in future sign in to the client without entering a password — the identity comes solely from the Windows handshake, there is no field into which a name could be written. Four barriers are deliberately set in the process: Only accounts linked to the directory are reachable — local accounts and in particular your emergency account remain outside SSO, since otherwise the convenient sign-in would remove precisely the password requirement that constitutes emergency access. What is matched, moreover, is the immutable identifier of the AD account, not the sign-in name: a name deleted and later reassigned would otherwise inherit the predecessor's permissions. Please note: The handshake itself could only be conclusively tested against a real domain controller — test the sign-in at one workstation before rolling it out across the board.
• NewA wizard converts existing local accounts to AD sign-in. During commissioning the first administrator account is created locally, before the directory is connected — afterwards the same person exists twice, and the directory reconciliation deliberately skips the local account. Nothing therefore happens by itself: the account remains, the AD sign-in never works. The new wizard under Settings › Directory converts the account without replacing it — roles, MFA devices and all references from tickets and logs remain attached to it; the wizard shows you the permissions before and after for comparison. Two things are important: After the conversion there is no fallback to the old local password — in the event of an AD disruption nobody would get in any more, which is why the wizard requires a second administrator account that can sign in locally and otherwise refuses the conversion. And the AD sign-in is actually carried out with your password before the conversion, not merely assumed.
• FixedWarnings, success messages and notices were too light for their white type. Six colour pairs missed the WCAG AA legibility standard — most clearly the blue notice with a contrast of 2.8 to 1 instead of the required 4.5. Affected were the warning colour, the success colour, the info colour as well as the two warranty tags in the device lists — in all six colour schemes, because these colours apply independently of the scheme. They are now one step darker; the hue stays the same, the type on them is legible. This becomes visible above all where you are looking closely anyway: at a warning, an expired warranty or a success message after saving.
• ImprovedThe automatic contrast check now looks at the status colours too. Ordivis Platform checks the legibility of its colour schemes itself on every build. Until now this check covered eleven colour pairs — controls, surfaces, navigation — and reliably reported „everything passes“. The warning, success and info colours of all things were not in it; and precisely there lay the six violations. 26 pairs are now checked, in addition the labelling of every status badge and the colour swatches of the dependency view. A checklist that inspects the frame and leaves out the warning notice reports success for the very case that matters.

2026.07.31.225131.07.2026

• ImprovedThe roadmap shows the actual state again. Under „Now“ and „Next“ stood topics that have long since been delivered — emergency management, ISMS, the operations and emergency manual, the checklists, the extended self-service portal, distributed inventory taking, platform monitoring, tenant administration, the BPMN editor and a good deal more. These eighteen points now stand in a section of their own called „Done“; under „Now“ stands only what is really being worked on. Two points are explicitly not carried over too: Multi-channel alerting today notifies only within the application — SMS, app push and voice call are stored as a contact list but not connected; and there is no graphical workflow designer. Both therefore remain shown as open.
• NewKnowledge articles can be looked up in the Android app without a network. Whatever you have opened once stays on the device. In the basement or the plant room the app then shows the remembered state — and says when it is from, so that a solution does not wrongly look like the current one. Only what you have opened is remembered, not the entire knowledge base: a lost phone should not be a copy of your internal documentation. On signing out and on deleting the local data, what was remembered goes with it.
• ImprovedThe ticket history in the app is coloured as in the client. Messages from the reporter are green, replies from the agents blue, internal notes amber, automatic entries grey and narrow — with an accent stripe, an icon and indentation. Until now all entries looked the same; who had written stood only in small type in the footer, and an internal note differed from a customer reply solely by a small badge. The colour never carries the statement alone: the icon in the header says the same thing.
• NewAll ticket kinds can be reported from the app, not just incidents. Incident, request, problem and change stand as a selection above the form, like the four options in the client. The incident remains preselected — it is the normal case on site, and anyone who merely wants to report something should not have to set anything. Without a network too: the remembered report keeps its kind.
• ImprovedThe app's start page now offers every route only once. „Report incident“ stood as a tile next to the button „Report“, „My tickets“ next to the entry „Tickets“ in the bottom bar, „Scan“ next to the viewfinder in the header, „Devices“ next to the bar entry of the same name. Two routes to the same destination are not convenience but the question of whether it is the same destination. The quick access now carries only what is not reachable anywhere else — inventory, knowledge and metrics — and disappears entirely if none of it is released.

2026.07.31.213031.07.2026

• NewNFC tags on the device — in addition to the label, not in its place. With the Android app you hold an NFC tag to the phone and assign it to a device; on assignment the same verified reference is written to it that also stands in the QR code. On the scan page the camera and NFC run at the same time — you do not choose a mode, you simply hold up whatever is there. Why both: A sticker is reliably readable on metal and in the cold, but must be visible and clean — in a closed cabinet, behind the device or under dust it is not. An NFC tag is read through plastic and in the dark, but fails immediately on metal. Anyone attaching both to hard-to-reach devices can get at every one of them when it matters. A device may carry several tags; a tag that already hangs elsewhere is not moved over silently but only after a confirmation.
• NewLabel printers from Zebra, Brother and Epson. Until now there was only the A4 sheet as a PDF. Ordivis Platform now additionally generates the print data stream for Zebra (ZPL II), Brother (P-touch/QL) and Epson (ESC/POS). With Zebra and Epson the Printers sets the barcode itself — it therefore appears at the resolution of the print head rather than that of an image, which decides legibility on small labels. Ordivis Platform does not print itself: You receive a file and pass it raw to the device (by copying it to the printer share, for example); no connection to devices on your network is established by itself.
• NewLabel and media library with a test print. Under Settings › Label printers & media you store what is loaded in your devices: dimensions, resolution, die-cut spacing, darkness — several media side by side, because an organisation often has a die-cut label on the warehouse printer and and a continuous tape on the handheld. A height of 0 explicitly means continuous material. The Test print also runs for a medium that you have not yet saved — the dimensions are rarely right at the first attempt, and this way your library does not fill up with failed tries.
• ImprovedWhat a printer cannot do, Ordivis Platform tells you — instead of silently printing something else. Not every device language masters everything: Epson does not know DataMatrix as standard (QR is set instead), and the Brother raster route has no dots above the letters, which is why umlauts are transliterated („Büro“ → „Buero“). Every such deviation appears as a note on the print job. A silent substitution would only come to your attention when scanning with the inventory reader — or never.
• ImprovedThe manual describes device labelling completely for the first time. New section in the CMDB chapter: stickers and NFC compared (including the frank statement that the identifier of an NFC chip no a security attribute), setting up the media, assigning and scanning. Please note: The output for Zebra, Brother and Epson follows the respective documented command set and is checked automatically — it has, however, not yet been accepted on a device from every manufacturer — please carry out a test print before labelling larger inventories.

2026.07.31.141331.07.2026

• FixedAn update could come to a halt halfway. The web portal was then already on the new version, the API still on the old one — and its service stayed stopped. The cause: when exchanging the API the installation helper ends the running server process. In doing so it inadvertently ended itself as well, because it runs as its child process. After that Ordivis Platform refused every new attempt for 30 minutes with the notice that an update was already running. The error did not occur with every update — only when the server process could not be ended by itself; that is precisely why it went unnoticed for a long time.
• ImprovedAn update brings its own installation helper with it. Until now an update was always carried out with the helper of the installed version. A correction to it therefore only took effect at the update after next. From now on the helper of the target version is used. A note on this very update: It is still applied by the previous version — the protection from the first point therefore only takes hold next time. If this update comes to a halt, start the service „Ordivis API“ and trigger it again; the lock releases by itself at the latest 30 minutes after the start.
• ImprovedThe message „An update is already running“ is now informative. It previously named only the start time. Anyone facing it could tell neither whether anything really was still running nor what to do. It now states when the lock releases by itself and where the file lies that holds the operation.
• FixedThe directory reconciliation with AD/LDAP could fail completely. A single user account sufficed: if it contains a binary attribute — a security identifier or a certificate, for example — that could wrongly be read as text, the entire run aborted. Affected tenants then received none at all directory data, in the log only a terse „LDAP sync failed“. Binary attributes are now skipped cleanly; the reconciliation runs through with all remaining fields.
• FixedA directory field could end up as „System.Byte[]“ in the display name. If an attribute came from the directory in binary form, Ordivis Platform took this technical placeholder over as the value — without an error message and without a log entry. Such values are now discarded; the field then stays empty instead of asserting something false.
• NewScheduled reports can be created — and really are sent. The button „New report“ now leads to a dialogue for schedule and recipients; the list shows the existing schedules, individual ones can be deleted. Above all, however, the stored schedule really is evaluated and the finished report sent to the recipients entered — until now neither happened.
• FixedA report definition without a query ended in a server error. Anyone saving a definition without a query got a meaningless error 500 — the check took place only in the database. An empty query is now permissible and is saved as „none yet“; a faulty one is named as such instead of letting the operation crash. Changing an existing definition was affected too.

2026.07.31.120231.07.2026

• FixedThe button „Import“ in the list of configuration items did nothing. It looked usable and triggered no operation whatsoever on being clicked — the worst of all responses, because it looks like a mistake by the user. The import now runs completely: first choose the CI class (the columns depend on it), then look at the structure of the file, then choose the file, then a dry run — and only after explicit consent is anything written. Lists without an import route now show the button greyed out, instead of making a promise they cannot keep.
• NewSample file with all fields — for download from the application. Under Settings › Import/export you choose the area (with the CMDB additionally the CI class) and download the template as CSV or XLSX. It brings the complete header row with it, a sample row to overwrite and — with XLSX — a second sheet explaining every column with type, mandatory flag and permitted values. The template is generated, not maintained: Its columns come from the same source the import reads from. Create a new attribute and it appears in it without your doing — a file maintained by hand would be out of date at the first new field, and nobody would have noticed.
• NewThe application now says what the file has to look like. Until now there was only a drop area without any information. Now it states column by column: name, type, mandatory, meaning, permitted values and an example — along with the rules for separators, dates and numbers. The area selection moreover names only the areas that really have a file import; previously two that have none stood there as well.
• FixedThe bulk import did not take over the class-dependent fields. Every object was created without them — everything that constitutes the CI class (IP address, operating system, location …) fell by the wayside wordlessly, and the row nevertheless counted as successfully imported. The columns now run through the same validators as the input form: an import no longer lets through what would be refused by hand.
• Fixed1234.56 became 123456. Numbers and dates were read according to the server's locale; a comma counted as a thousands separator. The same file therefore produced a different result depending on the server — with the purchase price a hundredfold. Also fixed: an unreadable date ended up silently as an empty field in the inventory, a mistyped column heading made the whole column disappear without trace, and the trial run always reported „0 imported“ — it could therefore not answer the one question it is supposed to answer.
• FixedA device's ticket list showed all of the tenant's tickets. Under the heading „Open incidents / changes for this CI“ stood all tickets — including those with no relation whatsoever to this device. That is more than a wrong list: the view asserts a relationship that does not exist. The cause was a filter that the query did not know and therefore discarded wordlessly. In addition the tab is called „Open tickets“ and now shows only those.

2026.07.31.101131.07.2026

• FixedAn image attached by e-mail was not visible on the ticket. Until now only images were stored that are embedded in the text the mail. A screenshot attached in the ordinary way — the normal case with an error report — was afterwards held exclusively in the archived raw mail: whoever worked on the ticket had to download it and open it in a mail program to see the very evidence at issue. Attachments now hang on the ticket, and displayable images additionally appear in the history. They are checked as before for size, file type and malware; whatever goes into quarantine deliberately appears not in the history. Applies equally to replies to existing tickets.
• NewImages in the history can be viewed enlarged. In the history an image stays small so that the text remains readable. A click opens it at original size — with zoom and scrolling, fitted to the window size, so that even a screenshot at full resolution remains manageable.
• ImprovedThe reporter of a ticket created by e-mail is the person, not the mailbox. If a report comes from another organisation's service desk, its system mailbox previously stood as the reporter on the ticket; the human being who noticed the error could only be found in the body text — a reply from the ticket did not reach them. Ordivis Platform now names the reporter in a header of its own that only Ordivis Platform sets itself and only where a human being wrote the text. The reply address alone was not enough for this: newsletters, distribution lists and other ticket systems routinely set one that does not belong to the reporter. The header therefore determines the reporter of a new ticket — who may write to an existing one is decided, unchanged, by the reply token alone.
• FixedThe setup files carried a shortened version number. Off 2026.07.31.0819 became in the file name 2026.7.31.819. That was not a naming error but the source: what was read was the numeric form of the Windows version resource, and there Windows stores a number per position — it cannot contain leading zeros at all. What is used now is the character string of the file version, which carries the number unchanged. Affected all three packages (main installer, wallboard, collector).
• ImprovedManual: a commitment put right. The administration manual continued to state that there was no function that transmits error reports to the manufacturer — „neither with nor without your consent“. Since the reporting function in the client that is no longer true, and two sections earlier the same manual describes exactly it. The paragraph now names both routes and what applies unchanged: no telemetry, no silent crash report, nothing in the background — every route outwards presupposes an action on your part.

2026.07.31.081931.07.2026

• FixedAn error report did not become a ticket at support. The mail went out and did arrive — it was merely discarded wordlessly at the recipient, while the program said „Your report has been transmitted“. The cause was two rules each correct in itself: Ordivis Platform marks every e-mail it generates as machine-generated so that an out-of-office reply does not trigger an endless loop — and a ticket system discards machine-generated e-mails for exactly the same reason. The error report is the one exception: a human being writes its text. It now goes out as such. This concerns you directly if you redirect the reports to a ticket mailbox of your own.
• ImprovedAn error report now says who it comes from. Until now it contained the tenant's identifier — a character string that nobody can do anything with. Now the report names Licensee and Tenant name, and as the reply address it carries the reporting person instead of the mailbox it went out through. A reply therefore reaches whoever noticed the error.

2026.07.31.063331.07.2026

• SecurityFree SQL queries in the reporting system could change data. The manual promises that such a query read only may. Nothing enforced that: the text entered ran against the database unchanged. Someone who is merely allowed to create reports could therefore overwrite or delete data. The boundary is now drawn by the database itself — the query runs in a read-only session. A mere keyword check would not have sufficed: PostgreSQL knows constructions that begin with SELECT and nevertheless delete. That is why this update is mandatory.
• SecurityReply e-mails were assigned without a tenant check. Ticket numbers are assigned consecutively per tenant and are therefore not unique between tenants. When assigning an incoming reply this condition was missing. In an installation with a shared database a reply could therefore end up on the identically numbered ticket of another tenant.
• FixedScheduled tasks reached only one tenant. By far the greater part of the background tasks worked exclusively on the database configured first — and thus passed every further tenant by: SLA deadlines and escalation, automatic closing, follow-ups, contract and licence deadlines, metrics, the checklist schedule. None of it failed; the tasks simply reported „successful“ without having done anything. Every task now runs one after another for every active tenant. If it hangs on one, the others are still processed and every cause reported individually.
• FixedThe cross-module search found nothing at all, as a matter of principle. All four sub-areas — configuration items, tickets, IP addresses, users — queried fields that do not exist under those names. Because every error was swallowed in the process, the search reported not a fault but „nothing found“. It now works; if a sub-area does fail in future, that appears in the log instead of passing as an empty result.
• FixedSLA attainment in the report measured the wrong thing. What was counted were tickets no escalation instead of tickets within the deadline — and that was off in both directions: a ticket escalated by hand or declared a major incident counted as a miss even if it was resolved on time. What counts now is the deadline alone. A target that elapsed during a maintenance window still does not count as a breach. Existing tickets and the stored trend curve were recalculated once — figures from older reports may therefore differ.
• NewEscalation levels now have a time axis — and you set it. Until now the check raised the level on every pass: an overdue ticket reached the highest level within three minutes and triggered three notifications to three recipients. Under Settings › SLA profiles you now determine for each level how many minutes after the deadline it takes hold (default 0, 60 and 240 minutes) — or switch escalation off entirely.
• NewThe audit log can be exported. The filtered view is saved as a CSV file — with a header naming the period and completeness. The separate permission audit.export is needed for this; the read permission alone is not enough, because an exported file leaves the system. Until now this permission stood in the catalogue without having any effect.
• NewHalf public holidays. On Christmas Eve and New Year's Eve many places are off duty from midday. Such days can now be entered in the service hours calendar — with the time from which the service hours end. Until then the SLA clock runs normally, after that it stands still. Municipal public holidays and company shutdown days that the federal state import does not know can be added in the same way.
• FixedA service hours calendar without time windows paralysed the creation of tickets. Anyone creating a calendar and intending to maintain the hours later got an incomprehensible error message from the date calculation at the next ticket. Such a calendar is now refused on saving, and the deadline calculation states clearly, in case of doubt, what to do. Also fixed: on a half public holiday the time after the end of business was consumed as service hours.
• FixedTickets created at the same time could receive the same number. The number was derived from the count of existing tickets; if two people created one at the same moment, one of the two failed with a server error. The number now comes from a counter in the database. As a consequence numbers can have gaps — a missing number therefore does not mean that a ticket has disappeared.
• FixedRestart order: the unresolved came first. If the dependencies of a system could not be determined, it counted as „depends on nothing“ — and therefore stood, of all places, first in the recovery plan. Such systems now come at the end and are marked as unresolved.
• ImprovedChangelog: mandatory update and minimum version at a glance. Both can be recognised on the entry itself as a road sign — a warning sign for a mandatory update, a stop sign for the oldest version still supported. A short legend at the top of the page explains the signs.
• ImprovedManual brought up to date. Newly described are the escalation levels, half public holidays, the audit log export including the content of an entry, the login lockout after failed attempts and the fact that only one login is valid at a time. In addition, 25 missing permissions in the permission matrix and 14 interface areas in the API reference were added.

2026.07.30.193330.07.2026

• FixedThe watchers of a ticket received no e-mails. Anyone who entered two colleagues as participants got nothing through to them. Three things came together: internal participants are recorded on the ticket with their user ID only, and their address was never looked up. Their sole route was the notification inside the application, and that sends no e-mail of its own accord. And status changes, handovers, appointments and attachments raised no event whatsoever for the watchers. Participants now receive the same e-mails as the reporter — internal notes remain excluded from this.
• ImprovedEvery ticket e-mail now contains the whole ticket. Until now it held a single sentence, truncated to 140 characters. Anyone receiving such a mail first had to log in to understand what it was about — and an external participant could not do that at all. The mail now carries the header data, the description, the public history and the list of attachments. Internal notes are never part of it.
• NewA new participant is brought up to speed at once. Anyone added to the watchers immediately receives an e-mail with the ticket's history so far — and does not have to ask what it is about.
• FixedAn uploaded file reached nobody. The upload triggered no notification whatsoever. That something did arrive occasionally was a side effect of the Windows client; anyone attaching a file through the portal or by e-mail produced nothing at all. An attachment is often the information the ticket turns on — the error picture, the log, the quotation. Just as silent until now was the cancellation of an appointment: only acceptances were reported.
• FixedThe decision-maker could not be reached. The field took a name and nothing else, and the button built an e-mail link on that name out of it — which could reach nobody. The decision-maker is now picked from the people search and receives an enquiry with the complete ticket at the push of a button. An entry typed in by hand remains permissible as a note; the interface then points out that no notification is possible.
• ImprovedStatus notifications stay manageable. Only changes that mean something to a reporter go outwards — in progress, waiting for customer, resolved, closed. Waiting internally for a service provider no longer produces any mail; the agent still sees all of it.
• NewReporting errors straight from the program. A bug icon in the status bar opens a form that sends the current page along as context. A data protection notice comes before the entry, and every piece of evidence can be deselected individually: today's log, a screenshot, system data. The screenshot is switched off by default and is displayed before it goes out. If the server cannot be reached — which is often exactly when one reports something — the client puts everything down as a file and opens an e-mail draft.
• Fixed„Saving failed: One or more validation errors occurred.“ This message said nothing. Out of the server's answer the client read, of all lines, the one with the standard sentence and discarded the statement of which field had been objected to. The cause of the error itself: after a dropped connection the ticket page carried on saving although it had loaded nothing at all, and sent empty fields — until the client was restarted. Both are fixed; connection errors now also name the cause and what to do.

2026.07.30.053130.07.2026

• FixedOn the right the process map did not show the processes of the tile that had been clicked. The title named the right tile — „53 utilities and waste management“, for instance — but below it stood the complete list of the selected administrative level. Affected was the middle band with the core processes, because its tiles stand for a product area and not for a single product. The tiles of the two other bands filtered correctly from the outset.
• ImprovedThe assignment of the processes itself was correct. Checked against an example: the product area „53 utilities and waste management“ carries exactly the 19 processes of the municipality/town level that its tile names — operating sewage plants, disposing of sewage sludge, keeping the sewer register and so on. It was a display that did not pass the filter on, and not an error in the catalogue.
• FixedIn the tree every level now narrows the selection, not just the product level. A click on a product area or a product group previously reset the selection silently: the detail card showed the node that had been clicked, the list beside it stayed unchanged. An area or a group now shows all processes beneath it; „clear selection“ takes you back to the full list.
• Fixed„Adopt into your own set“ would have created far too much when an area was selected. The same missing filter took effect when writing: the notice announced the 19 processes of the area, whereas the 1,718 processes of the entire administrative level would have been created. The adoption now stays inside the branch you can see.
• ImprovedThe database reference in the manual is complete again. It is generated from the stored data model, and that model did not know 31 tables — which is why they were missing from the reference although they are created in every installation. Added are, among others, the state law citations and the collector's client inventory.
• ImprovedPrecaution against a faulty database update. Because the stored data model lagged behind the actual one, the next schema change would have written foreign tables along with it — exactly the sequence that made an update unusable on 16.07. The model is now reconciled against the real state, with evidence for each module. This release therefore changes nothing in your installation's schema.
• ImprovedThe installation package of the discovery collector is 43 MB smaller (110 instead of 153 MB). It dragged the server substructure along with it — PDF generation, certificate management, database access — which a scanning machine on site never touches. The four scanners for WMI, SSH, SNMP and LDAP now stand on their own; nothing changes functionally.

2026.07.30.002230.07.2026

• FixedThe process catalogue cited the state law of Hesse — outside Hesse as well. In 1,318 of its 5,019 processes there stood a citation from the state law of Hesse: HGO, HBO, HSchG, HSOG and 28 further abbreviations. For a municipality in Bavaria „§ 66 HGO“ is not merely unsuitable but wrong — and nowhere did it say so. The remaining three quarters or so cite federal law (SGB, WHG, BauGB …) and apply unchanged in every state.
• NewThe federal state on the tenant — with a suggestion, but nothing automatic. Ordivis Platform derives a suggestion from the postcode in the tenant profile, and you confirm it. Why not automatically: postcode districts follow logistics, not state borders. Where two states share a district, Ordivis Platform says „not determined“ in so many words and suggests nothing at all. A plausible-looking wrong suggestion would do more harm here than none, because it ends up as a legal basis.
• NewThe citation for your state on the process. If the federal state is recorded, the detail card of a catalogue process shows, next to the source, the act that applies where you are — with its official title and abbreviation. Supplied are 16 areas of law for all 16 states: municipal constitution, district and municipal election law, building regulations, school law, police and public order law, road law, fire and disaster protection, state civil service law, archive law, data protection, administrative procedure, water law, monument protection, the public health service and the emergency medical service. That covers 1,169 of the 1,318 affected processes.
• NewVerified and suggested are kept apart. Of the citations supplied, those on municipal constitution and archive law are reconciled against a source. All the others are reasoned suggestions and are marked as such, with the request to look them up on the state law portal. An unverified statement that looks like a verified one would be the worse error.
• ImprovedAs long as the federal state is missing, no citation is asserted. The detail card then says that the text comes from a Hessian edition and what to do about it. The original text of the source stays in place in every case — it is part of the catalogue as published and has to be carried along under the licence. Anyone who can no longer see the origin cannot judge a deviation.
• ImprovedThe three city states do not get an invented municipal code. Berlin, Bremen and Hamburg have none, because the municipal and the state level coincide. A note to that effect stands there, together with the relevant state constitution.
• ImprovedProcess catalogue and product catalogue checked against each other. 5,019 processes against 148 products: no orphaned assignment, the product area matches the product number throughout, exactly one product area per product, all process names unique. 25 products carry no process — they are the operating and funding products (theatres, museums, libraries, hospitals, sports facilities, utilities and waste management, public transport, economic development). Two deviating product names in the source file have no effect: what counts is the product number, the name comes from the product plan.
• ImprovedThe manual states the limits in figures. The chapter on organisation now says how many processes depend on the state, which areas of law are covered, which 15 are not yet, and why the federal state has to be confirmed.

2026.07.29.231329.07.2026

• NewDiscovery collector for remote networks. A small Windows service that you install in the other network — in a branch office, a sealed-off production segment or at the customer you look after. It searches the network ranges assigned to it and reports its findings to the central system. Until now discovery ended at the first network boundary: the scans ran on the server, and what you cannot reach you cannot inventory.
• NewThe connection always starts at the site. The collector calls the central system, never the other way round. No inbound firewall rule is therefore needed at the remote site, and a site behind a NAT router works exactly like one with a fixed address.
• SecurityEnrolment by way of a voucher that can be redeemed once. You create it in the collector administration, the collector redeems it at first start and receives its permanent access. After that the voucher is spent — a second machine cannot use it again. Blocking and renewing revoke the access immediately, regardless of whether the machine at the site is still running.
• SecurityThe scan password does not leave the site. It is encrypted on the machine at the site with the Windows machine key; the central system never learns it. Otherwise it could lose it across every site at once. A plain-text password in the configuration file is rejected and reported, not used quietly.
• NewA dropped connection is not a loss of data. If the central system cannot be reached, the collector puts its report into a local queue and hands it in at the next contact. Because every report carries an identifier of its own, the central system recognises the latecomer and does not book it twice.
• NewClient inventory: the machines found and their applications. In table form like the IPAM — filter by site and review status, search, export as CSV or PDF. Alongside it an application view with one row per application and version and the number of clients affected: the question that actually gets asked in operations — where is this running?
• NewAdoption into the CMDB only after confirmation, with a duplicate check beforehand. Before creating anything, Ordivis Platform shows existing configuration items with the same MAC address or the same host name; you decide whether the finding belongs to one of them or becomes a new entry. Anyone skipping the check creates duplicates and only notices when both have to be maintained. A discarded finding does not come up again at the next run.
• NewOperating state per site. The collector list shows enrolment pending, active, active (runs off), overdue or blocked — together with the last sign of life, the last run, the number of findings and the open ones. A collector only becomes overdue once twice its reporting interval has elapsed (at least one hour): a single missed sign of life is not yet an incident.
• NewAn installation package of its own for the machine at the site (46 MB instead of 675 MB). It contains the collector service and nothing else. Sending the complete package to a branch office or a customer would mean supplying the server, the database and the portal along with it — software that has no business on a scanning machine.
• ImprovedWallboard and collector can be selected individually in the main installer. Both are additional roles, not installation types: as an installation type of their own if the machine is meant to be nothing else, or as a single tick alongside a server or client installation. Anyone needing just one of the two selects exactly that.
• NewDNS/PTR reconciliation for the IPAM. A connected tool reports forward and reverse resolution; from that Ordivis Platform sets the DNS name on addresses already known. Contradictions between the two directions, and entries pointing to an unknown address, are reported — not cleaned up. Deleting a DNS record is a decision with consequences and does not belong in a reconciliation run.
• NewThe site from AD Sites & Services instead of guessed. The site stated on a subnet is adopted if a site of that name exists in the CMDB; if it does not, the name appears in the receipt — so you can see which site is still missing. An assignment already in place is never overwritten: a scan does not replace a person's decision.
• NewAD computers, software inventory and network devices as candidates. External tools such as the free Ordivis Toolbox can deliver them; they land in the same review queue as the collector's findings. There is therefore exactly one place where findings are reviewed — no matter where they come from. LAPS passwords and BitLocker recovery keys are never transferred in the process.
• NewCertificate inventory with expiry monitoring. Reported certificates arise as an entry of the certificate class in the staging state and expressly not verified — visible and monitored, but recognisably unconfirmed. They are identified by their fingerprint: the same subject after a renewal is a different certificate.
• NewVulnerability findings become assessed risks. Reported CVE and WID findings land in a risk analysis of their own for the information domain, where they are given an assessment and an owner. A separate list of findings would be a list without effect. Only the reference is transferred — identifier, component, CVSS, time of discovery — and no exploit details. If the CVSS score is missing, „medium“ applies: a finding without an assessment is unknown, not harmless.
• ImprovedAn integration token gets its permissions when it is issued — and keeps them. The new feed-in areas do not reach existing tokens; a new one is issued for them. What a machine account may do is decided deliberately and does not accrue to it after the fact.
• ImprovedThe manual updated — and three wrong permission names corrected. The discovery chapter named permissions that do not exist in the system at all. Newly described are the collector, the client inventory, the three installation packages, the machine-facing seam of the interface and the risk analyses that get filled.

2026.07.29.202429.07.2026

• NewOZG service catalogue for public administrations. The implementation catalogue under the Online Access Act stands as a fourth pillar beside the product plan, the process catalogue and the business allocation plan: 18 topic areas, 54 life and business situations, 575 services. It shows the same administration from the outside — ordered by the situation in which people and businesses need a service. Visible only for tenants whose organisation type is public administration.
• NewEvery OZG service finds the processes that belong to it. The link runs through the LeiKa key at product level — it points to one product, but to as many as 722 individual processes. Lead responsibility follows from the product assignment: whoever is responsible for the product in the business allocation plan is responsible for the OZG service as well. Derived assignments can be added to and, with a reason given, revoked.
• NewState of implementation and state of digitalisation. For each service the OZG maturity level, the kind of implementation (in-house development, EfA reuse, EfA provision, portal network), the specialist procedure and the online service are recorded. The state of digitalisation sums that up per topic area and organisational unit. „Not yet assessed“ remains a state of its own and is not counted as maturity level 0 — otherwise the picture would systematically look too good.
• NewNationally uniform texts in the product, process and OZG catalogue. Full text, required documents, deadlines, costs and legal bases from the federal FIM catalogues supplement the previously terse catalogue entries — supplied with the product, usable without internet access.
• NewReconciliation against the federal master processes. On request the reference sequence modelled by the federal government is fetched from the FIM process library and compared with your own released process definition. A missing legal basis is a gap, a missing step merely a note — and additional steps of your own are not a finding: the master process shows the legal position, not the way a house is organised.
• ImprovedThe implementing level is shown as precisely as the source allows. „State or municipality“ stays „state or municipality“ — which level is responsible is decided by state law. The filter „Only what concerns us“ therefore hides only what demonstrably concerns another level.
• FixedAn OZG catalogue that had not loaded looked like an empty one. If the server did not answer — because it is older than the workstation program, for instance, and does not know the interface yet — the page showed an empty catalogue without saying so. Anyone expecting 575 services and seeing nothing looks for the error at their own end. It now says that nothing was delivered, and why.
• FixedThe OZG audit report announced „no gaps“ when it could not be retrieved at all. A report that sounds the all-clear after a failed retrieval is worse than none. It now says that it says nothing.
• ImprovedThe manual and the in-app help extended by the chapter on the OZG service catalogue.

2026.07.29.152029.07.2026

• SecurityLabels carry a check value. The QR code of a device label now additionally contains a short check value from which the server recognises that the sticker comes from this Ordivis installation. Without it, a label could be printed that points to someone else's device — whoever sticks it on leads the technician to the wrong profile when scanning, and the incident report ends up on the wrong device. The check happens on the server, not on the phone. Labels already printed without a check value remain valid.
• ImprovedThe manual and the in-app help brought up to this version. The product version of the manuals was two releases behind; the PDFs and the help embedded in the workstation program have been rebuilt.
• FixedThe history on this page was four releases behind. The releases of 28 and 29 July were in the app but missing from the public history page. They have been added.

2026.07.29.095729.07.2026

• FixedOnly one update run happens at a time now. If the auto-update rule started in the maintenance window while a run triggered by hand was already under way, the two collided: the second aborted because the first was in the middle of exchanging the program files, and in doing so overwrote the shared status file. The progress display then announced „update failed, previous version restored“ although the update had succeeded. A second run now stands back and leaves the running one untouched.
• FixedA cancelled download is no longer reported as „file missing on the update server or its checksum does not match“. Both cases ran into the same message until now — and sent the search for the fault in the wrong direction when in truth only one's own service had been shut down.
• ImprovedIf an update run hangs without reporting its end, after 30 minutes it counts as stale and no longer blocks future updates. An update blocked for good would be worse than a duplicate run.
• ImprovedThe database reference in the manual extended by the two new fields on the organisational role (deputy as a name and as a reference to a person configuration item).

2026.07.29.081629.07.2026

• NewOrganisational roles carry a deputy. Holder and deputy sit in one entry, because an officer function is appointed rather than filled: „information security officer: Meier, deputy: Schulz“ is a single statement. Split across two records it would only raise the question which of the two is the role — and the business allocation plan would show one function twice. The same person as their own deputy is rejected.
• NewSuggested people in every person field — when filling a post, and for the holder and the deputy of a role. From the second letter onwards suggestions appear from three sources, each with its origin: user accounts, person configuration items and the names that already occur in this organisational structure. The third source is the normal case: an organisation just imported has neither accounts nor CIs for its staff. Free entry remains expressly permitted — a new colleague is in the organisation chart on day one and does not have an account yet.
• NewThe CSV exchange of the organisational structure carries the deputy of a role along with it (column „rollenvertretung“). Without it the new field would have been lost on the round trip of export, Excel, import.
• FixedIn „My day“, resolved and closed tickets stayed put under „My appointments“. The status filter hung on the due-date branch alone; follow-ups and deferrals went past it. The reminder job has always filtered across all three dates — both views now mean the same ticket. The date itself stays on the ticket: that a follow-up had been set is part of the history.
• ImprovedThe role list shows holder and deputy in one row, and a role nobody holds says so.
• ImprovedThe organisation chapter of the manual extended by the deputy on roles and the entry of people, in-app help included.

2026.07.29.063429.07.2026

• NewA new menu area, Organisation, with four pages. Process organisation and organisational structure are open to every tenant — a process map and an organisation chart make just as much sense for a company. The product catalogue and the business allocation plan appear only for the organisation type public administration. The area becomes visible through the new permission gvp.read; maintaining it requires gvp.manage.
• NewReference catalogue for public administration. The municipal product plan of the KGSt (6 main areas, 16 product areas, 41 product groups, 148 products) and 5,019 administrative processes across 462 specialist tasks are available without any import. The catalogue is a reference and creates no configuration items: 5,019 processes as CIs per tenant would mean 5,019 CI numbers, audit rows and search hits for material of which a municipality never carries out about a third. Only the adoption creates business processes — repeatable, with a preview, without duplication. The source notice under CC BY 3.0 DE appears everywhere catalogue data becomes visible, and sits as a date on the catalogue edition instead of being hard-wired into the program.
• NewThe organisational structure with full maintenance. Organisational units can be created, changed, moved and deleted; below them posts with a post identifier and a scope, appointments with a share and a deputy, and officer functions. When moving a unit, a cycle check prevents upside-down trees; deletion is blocked as long as sub-units, posts or product assignments hang from it. To get started, the organisational structure of a model municipality can be imported — the posts arise unfilled.
• NewThe head of an organisational unit is now a post. The previous field for the head was a reference to a person on the record and could depict neither a vacant nor a shared nor a deputised headship. Existing entries are transferred automatically at first start into a head post with an appointment.
• NewImport and export of the organisational structure as CSV. A flat table, one row per post — the form in which an administration keeps its organisation anyway. The export delivers exactly the columns the import expects: download, add to it in Excel, load it back. A preview before writing, rejected rows are reported instead of swallowed, a second run changes nothing.
• NewProcess map and process tree. Three bands — management, core, support — with tiles and counts, drill-down as far as the process list, filters by administrative level and process type, and export as PNG. The administrative level is preset to municipality/town; without a filter it would be 5,019 processes and therefore no map at all.
• NewThe business allocation plan at the push of a button. Lead responsibility follows from a single assignment of a product to an organisational unit — when the head of a department changes, it is the organisation chart that gets touched, not every process. An explicit entry on the process trumps the derivation, and the origin is stated in the list, in the detail view and in the PDF: an inherited A must not look like a maintained one. Support, consultation and information, conversely, are inherited downwards and can be assigned at product, specialist task or process level.
• NewAn audit report and frozen versions. The report names processes without a lead, duplicate leads, products without a unit, units without a head and orphaned assignments. A version freezes the state of the data, the PDF and its SHA-256 checksum together with the date of issue and the author; a query for a given date reads the version instead of reconstructing the state. Two versions can be compared.
• ImprovedThe organisation chart no longer cuts labels off. The width of a box follows the longest label and the text is wrapped; at a fixed width a name such as Department 1 General Administration and Family ran into the neighbouring node.
• ImprovedWider and adjustable working areas. In the process organisation and the product catalogue the boundary between navigation and list can be dragged; a double click restores the original width.
• ImprovedA manual chapter on organisation and the matching in-app help for all four new pages.

2026.07.28.191228.07.2026

• SecurityWallboard tokens could obtain someone else's permissions. Anyone allowed to manage display devices could issue a token for any account of the tenant — including a more highly privileged one. Since such a token is exchanged anonymously and without a second factor for an access token carrying the permissions of the account it is bound to, an ostensible „wallboard“ was a route to someone else's permissions without a password and without a logged role assignment. The same barrier as for granting roles now applies: only a token for an account whose permissions you hold in full yourself can be issued.
• SecurityThe report preview executed free SQL with nothing but read permission. Such SQL runs unchanged against the tenant database and therefore past every module permission — anyone allowed only to view reports could get at identity data, log data and other people's ticket data this way. The SQL route now requires the permission to create reports; the building-block route stays with read permission. Without that permission the client hides the SQL switch instead of running into a rejection.
• NewDeferral and follow-up are separate dates. The deferral („deferred until“) hangs on the waiting status and is lifted when that status is left; the personal follow-up is untouched by it. Both carry a date and a time, and both stay visible after the reminder instead of being deleted. The ticket list has columns of its own for both, coloured by the next date, plus the view „Dates only“.
• NewThe resolution target can be set by hand. The sidebar carries a date and time picker that saves immediately. A date set by hand is recorded as such and no longer moves along with the duration of a waiting status — a commitment made to the reporter should keep what it says.
• NewReminders for all three dates. The job now covers follow-up, deferral and due date, each with an advance warning 24 hours beforehand and a notification to the agent when it falls due — each can be switched off individually through the notification catalogue. The SLA escalation to the group is untouched by this.
• Improved„My day“ is now called „My appointments“ and shows all three dates side by side. The due date appears there only when it is within view — overdue or within the next 24 hours. Otherwise every open ticket with an SLA deadline would be standing there.
• FixedThe wallboard could not be selected in the installer. The component was there, but out of reach — the wizard skipped the component selection and hard-wired the three known installation types. There is now a fourth installation type, „Wallboard“, for pure display devices, and an additional tick for server or workstation machines with a wall screen attached. Beyond that, the Windows runtime required is now set up for a pure wallboard installation as well — without it the device would not have started after the installation.
• ImprovedResponsibility on a process is recorded in one place only. „Accountable (A)“ and „Responsible (R)“ could be maintained with equal standing in the profile and in the RACI matrix, and in practice the two drifted apart. The matrix is now the source; entries already in the profile are adopted once, existing matrix entries stay untouched. The roles also carry German labels now.
• ImprovedThe dependencies view calls objects by their name. The page title was hard-wired in English and is now the German „Abhängigkeiten“; the list of connections shows the counterpart by its plain name instead of its internal identifier — and does so relative to the selected node. For incoming connections the selected node itself used to stand there.
• ImprovedThe workstation client starts maximised. Can be switched off under Settings ▸ Appearance ▸ Window.

2026.07.27.191427.07.2026

• NewWallboard: key figures on the screen in the room. A full-screen program of its own for display devices shows the key figures from the service desk, the CMDB and IP management in rotation — with large numbers, trend lines for the last two weeks, the agents signed in, a large clock and the tenant's logo. If a new ticket is waiting, the tile fades quietly in and out instead of flashing. The wallboard is a component of the installer that can be deselected and, if you wish, starts on its own when you sign in.
• SecurityDisplay devices without a password on the device. Instead of storing a user name and password on an unattended machine, every display device gets a named token of its own with a lifetime. It is bound to a user account — the screen sees exactly what that account is allowed to see — and can be revoked individually. On the device it is stored encrypted (Windows DPAPI).
• ImprovedEvery access for devices and programs on one page. Under Settings → Platform → Integration tokens now holds both kinds of machine access side by side — external programs such as Ordivis Toolbox and display devices — each with an explanation of its own about what it is meant for and how the two differ.
• FixedThe workstation update service is supplied again. The service that updates workstations automatically was missing from the release package and was not compiled along with it when publishing.
• ImprovedManual: wallboard and tokens. New sections describe the installation, the setup and the operation of the wallboard as well as both kinds of token with their differences and a recommendation on the account to use.

Week 30 · 20–26 July 202618 versions

2026.07.26.211526.07.2026

• SecuritySigning in with two-factor authentication is possible again. Anyone who had set up the second factor got in with no access at all any more: the confirmation code entered never reached the server, so the sign-in kept asking for it again endlessly. The code is now taken over and checked. This is why this update is mandatory — on an older version, accounts with a second factor stay locked out.
• NewThe confirmation code is asked for everywhere. The workstation program, the web portal and the Android app show a field for the 6-digit code from the authenticator app after the user name and password. In the workstation program the password and the tenant stay in place; a wrong code clears the code only.
• FixedA wrong password is reported as such again. A rejected sign-in led to an unspecific error message instead of a clear pointer to the password.
• FixedThe dependency browser and the business service map stayed empty. Both views need a starting element; opened from the menu it was missing and the page stayed without content. Both now have a search field at the top for choosing the starting element. On top of that, in the dependency browser the filters by level and by type now take effect together instead of cancelling each other out.
• ImprovedThe device assignment appears on the configuration item automatically. „Assigned to“ on the device and „Assigned devices“ on the user account are derived from the active asset assignment and can no longer be maintained by hand — so the two can no longer drift apart.
• ImprovedManual: signing in with a second factor. A new section describes the two-step sequence in the workstation program, the portal and the app.

2026.07.25.175925.07.2026

• NewA report builder with drag & drop. Reports can now be assembled by drag & drop from the fields of the module in question — with filters, sorting, grouping and key figures (count, sum, average) as well as a live preview. Specialists can still work with an SQL query.
• New69 ready-made reports for management. Every module brings at least five frequently needed standard analyses with it (e. g. open tickets by priority, expiring contracts, configuration items per class) — ready to use, exportable and sendable on a schedule.Addendum: „sendable on a schedule“ did not apply at that point. The schedule stored was not evaluated and nothing went out to the recipients; both only arrived with 2026.07.31.1413.
• NewA header and footer you can adapt. The header and footer text of the reports can be adapted per tenant; every report carries the date it was created. The mark „Created by Ordivis Platform“ is retained.
• ImprovedA new manual section & database reference. A section of its own in the manual explains reports and SQL queries and documents the complete database structure with field descriptions.

2026.07.25.152725.07.2026

• NewImmediate measures in the crisis dashboard as a board. The running checklists of an incident now appear in the emergency dashboard as a clear board (open / in progress / done). A click on a card opens the checklist to work through straight away — without leaving the dashboard.
• ImprovedSorting and filtering checklists by area. Templates can be assigned to one of eight areas (e. g. human resources, IT operations, business continuity) and filtered by area in the overview.
• NewTriggering checklists automatically — on a schedule or from a ticket. A template can now start automatically on a recurring basis (daily, weekly or monthly) or when a new ticket arrives — in addition to the existing triggers „crisis“ and „approved change“. The trigger is set explicitly for each template.

2026.07.24.175124.07.2026

• NewEncrypted database backups. The daily backup can now be stored with a central password as an AES-256-encrypted 7z archive file. You set the password in the backup settings; it is never stored in plain text (Windows DPAPI) and is asked for when restoring. Without a password it stays, as before, an unencrypted backup.
• ImprovedChecklists in the new area „Self-organisation“. The checklists and playbooks are now bundled in a menu area of their own; the former duplicate checklist view is gone.

2026.07.24.115424.07.2026

• NewProcesses in considerably more detail. On every business process, legal basis, scope, version and validity, recipient of the service, decision criteria, exceptions and error paths, media breaks, statutory deadlines as well as fees can now be recorded.
• NewProcess overview. A complete list of all processes with their fields in table form — searchable, filterable, sortable and exportable as CSV or PDF. A double click opens the process in the documentation.
• NewFields for public administration. For tenants with the organisation type „public administration“ additional fields appear: LeiKa ID, FIM ID, KGSt product number, OZG topic area and ID, maturity level, eIDAS level of assurance, federal level and administrative act (§ 35 VwVfG, German Administrative Procedure Act).
• ImprovedA menu area of its own, „Processes“. It bundles the new process overview and the (renamed) process documentation.
• ImprovedProcess import extended. The CSV and JSON import supports the new fields, with updated templates.

2026.07.24.071824.07.2026

• NewChange management considerably extended. A change request (RFC) in draft can now be edited, with a planned start and end scheduled, and linked with the triggering tickets and problems as well as the devices and services affected (CIs) — through a search field, without typing in numbers.
• NewApproval can be configured — for one-person operation too. By default the applicant may not approve their own change (four-eyes principle). In very small teams this can now be switched: if One-person operation is active, the same person answers for the request and the approval.
• FixedA new RFC was approved on the spot. The default for the type of change was „standard“ — and standard changes are pre-authorised. The default is now „normal“ and goes through the approval; a note explains the types. The statuses now carry German names throughout.
• NewImporting processes from a file. Business processes (profiles) can be created and updated in one step from CSV or JSON files. The import shows the file structure expected, offers templates to download and a preview (a dry run) before it writes.
• ImprovedOne dashboard instead of two. Key figures and work lists (overdue and new tickets, the spread of priorities) now stand together in the service desk dashboard; the separate menu item „Work lists“ is gone. A click on a key figure tile opens the matching filtered ticket list.
• ImprovedLists as clear tables. The lists on the dashboards appear as compact tables with column headings instead of stacked text.
• FixedKey figures slightly cut off at the bottom edge on all dashboards.

2026.07.22.205322.07.2026

• FixedA client update deleted the workstation's server address — and thereby prevented every further one. The address a workstation uses to connect to the server is stored per machine during setup. When the files were exchanged it was destroyed in two ways: one file is not part of the delivery and was removed as „surplus“, the other is part of it and got the factory setting localhost written over it. After that the update service looked for the server on the machine itself, found nothing there and had no effect — every update made the next one impossible. Settings specific to an installation are now neither deleted nor overwritten during the exchange, just as already applied to server settings and key files.
• ImprovedSecured against a relapse. The case is backed by a test that reproduces an update with a changed factory setting and checks that the address entered still stands unchanged afterwards.

2026.07.22.202522.07.2026

• FixedWorkstations on a machine of their own were never updated — the service asked the wrong server. On a separated installation the update service ran endlessly without effect: no download, no prompt, not a stir in the task manager. The reason: the server address is kept in three places — the connection set in the program, the entry from the setup and a default supplied with the product. The service did not know the one set in the program and fell back to the default localhost instead, while the program had long been talking to the right server. It now evaluates the same sources in the same order as the program itself. With that, workstations on machines of their own update themselves reliably for the first time.
• ImprovedYou can now see what the update is waiting for. After you agree, the service waits for Ordivis Platform to close — until now it did so for up to two minutes without any sign of life and could not be told apart from a hang. This step is now visible as a state of its own. If Ordivis Platform cannot be closed, a clear request to close it appears instead of a silent abort; the update then carries on by itself.
• ImprovedAn explicit „update now“ no longer waits two minutes. After a restart of the service a warm-up period ran that left even an agreement already given lying there. It is now cut short as soon as an answer is available.
• ImprovedMore meaningful log. The service now records which server address it uses and which source it comes from. Until now only the address stood there — which made it impossible to tell whether the server cannot be reached or the service is using the wrong address.

2026.07.22.192722.07.2026

• ImprovedDashboards show considerably more in the same space. Until now a few large tiles with one number each filled half the screen. Every area now has the same structure: a narrow bar with all the key figures at the top, charts below it, work lists below those. A click on a key figure opens the matching filtered list — „SLA breached“, for instance, exactly the tickets concerned. The period under review can be set to 7, 30 or 90 days.
• NewTrend charts for throughput and SLA attainment. How many tickets come in, how many are resolved, how the adherence to the times promised develops — until now that could only be seen as a snapshot. The daily figures are now additionally filled in retrospectively for 90 days, instead of letting the history start only from the day it was built in. As long as too few days are available, the chart says so — instead of drawing a line out of a single point.
• NewDistributions at a glance. The backlog by age, open tickets by priority, the CI lifecycle as well as protection requirements and the risk situation appear as a doughnut chart with figures in the legend. In the network area a bar for its occupancy stands behind every network.
• FixedTwo different numbers for „open tickets“ on the same screen. The tile counted only new and waiting tickets — that is without those in progress —, while the analysis beside it reported everything unfinished. On a test data set 18 and 36 thus stood next to each other. What counts now, uniformly, is everything that is neither resolved nor closed; „New“ stands beside it as a tile of its own.
• FixedA negative mean resolution time. If tickets carried a time of resolution earlier than their time of creation, the mean counted that in and reported an impossible duration. Such values are no longer put out as a figure.
• NewWork lists in the service desk. Under ITSM ▸ Work lists you will find the overdue and the new tickets as well as the spread by priority — the way into the daily work, kept apart from the analysis.
• FixedPlaceholders instead of characters in several texts. In a few places what appeared literally was <1d instead of <1d, the same for „greater/less than“ statements about deadlines and for quotation marks in note texts.
• ImprovedUpdate centre tidied up. The list „Components affected“ now stands directly beneath the version information instead of at the foot of the page, and its columns follow the content — longer names used to be cut off.

2026.07.22.161522.07.2026

• FixedWorkstation clients were never updated — the cause lay on the server. On separated installations (the client on a PC of its own, the server elsewhere) the workstation stayed on its version: the API and the portal were updated, the client was closed and started again just as old as before. The route intended for it could not work at all — the endpoints through which the server hands the files to the workstation demanded a sign-in that the updating process, by its nature, does not bring with it. Every request was rejected; as a substitute the product website was tried quietly, which came to nothing without internet access. Both are fixed.
• NewAn update service of its own on every workstation. A client installation now sets up a Windows service that keeps the workstation at the server's level automatically — on a schedule (every four hours by default) or at the push of a button. Without administrator rights: Until now every update called for a Windows prompt for elevated rights; anyone without them — that is, most users — could not update their client at all.
• NewYou decide when Ordivis Platform is closed. The service first downloads the new version in full and checks it — only then does it ask: update now or postpone by one, four or sixteen hours. Nothing is closed without your agreement. If Ordivis Platform is not open anyway, the update runs unnoticed in the background. After the exchange the service checks against the program file whether the new version is really in place; if it is not, it restores the previous state.
• FixedThe update centre showed the new version for three components although the workstation stayed old. The list describes exclusively components of the server — including the row „Client“, because what that means is the copy of the client next to the server application, not your own machine. The two now stand apart and labelled: „This workstation“ and beneath it the server components.
• FixedWorkstations ran ahead of their own server. The target used was the latest version published on the product website instead of the version the site's own server actually runs. A workstation could thereby have jumped to a state its backend does not know at all yet. What counts now is always your own server.
• NewNetwork capture: more details per device, feedback to IP management. The capture from Ordivis Toolbox now delivers additional characteristics per host, knows the manufacturer from the hardware address even without internet access and records where each address comes from. New on top of that is the return channel: network ranges detected and addresses not yet confirmed travel into IP management, and the lease lists of a DHCP server can be read in directly.

2026.07.22.125422.07.2026

• NewService hours can finally be maintained — public holidays per federal state included. The SLA clock has always reckoned with service hours calendars: time windows per weekday, time zone, public holidays. None of it could be edited until now — there were only the two calendars from the initial installation. Under Settings ▸ SLA profiles you now create calendars of your own, maintain the times per weekday and enter the statutory public holidays of a federal state for a year at the push of a button. The dates are calculated and not fetched — no internet access is needed for it. A repeated import replaces the year instead of duplicating it; the number stands in the heading.
• ImprovedProtection against deleting service hours that are in use. A calendar an SLA profile refers to cannot be deleted; the message names the profiles concerned. Without that protection a profile without a calendar would quietly reckon around the clock — and thereby tighten every deadline without anyone noticing.
• FixedThe misleading card „Global service hours“ removed. It offered a switch and a start and end, but was neither loaded nor saved: changes were gone again by the next time it was opened. Nor did it fit the model — service hours hang on the calendar of the SLA profile in question, and a global start and end does not exist there. The new calendar administration takes its place.
• FixedError messages without quotation marks. Messages from the server appeared in many places set in quotation marks, because the text was passed through unprocessed.

2026.07.22.110022.07.2026

• FixedStatus colours are displayed again. In the ticket list and in the legend, „In progress“ and „Waiting for customer / service provider“ appeared grey instead of yellow and orange. The cause was the way those two colours were stored: they were looked up again while drawing, which failed in overlaid windows — the legend, for instance. The colours are now stored firmly and can no longer be lost. It applies throughout: blue = new, yellow = in progress, orange = waiting for customer or service provider, red = blocked, green = resolved, grey = closed.
• ImprovedThe column widths of the ticket list follow the content. Ticket numbers and status markings used to be cut off because every column had a fixed width. They now adapt; the subject still fills the space that remains, and reporter, supporter and tenant have an upper limit so that a single long name does not crowd out the subject. Dragging by hand remains possible.
• Fixed„Save“ writes the AD connection only when something has changed. The save button in the settings rewrote the AD/LDAP configuration at every click — no matter which area was open at the time — and reported accordingly „AD/LDAP configuration saved“, even when you were working on entirely different settings. It is now only saved when something has actually changed.

2026.07.22.100022.07.2026

• Fixed„Update now“ really does update the workstation now. The notice about an outdated workstation client appeared correctly, but its button merely switched to the settings — where the update centre reported „you are using the latest version“, because that statement refers to the Server itself. The self-update was thereby out of reach in exactly the situation it was built for: server current, workstation behind. The button now starts the update straight away, and it is available in the update centre as well. If the start fails — because the update service is missing or the elevation of rights is refused, for instance — that is reported explicitly instead of quietly doing nothing.
• ImprovedThe update centre distinguishes between server and workstation. The page showed details of two machines side by side without making that plain: the version information and „Components affected“ describe the server, the status bar describes your own workstation. If the workstation is behind, the status display now says so explicitly („server current — this workstation outdated“) and names both versions in plain words.

2026.07.22.084422.07.2026

• FixedThe follow-up was not saved — and blocked the whole sidebar. A follow-up date once chosen did appear in the journal, but was gone by the next time the ticket was opened. The cause was the time zone of the date, which was rejected on saving — and rejected for the whole ticket: as long as a date was set, category, effort, group and decision-maker could no longer be saved either. The notice about it stood at the top of the page and remained invisible while working in the sidebar.
• FixedForwarded e-mails carry the right reporter. If a member of staff forwards a message to the ticket system and enters the customer as „reply to“, that customer is now taken as the reporter — until now the member of staff doing the forwarding was entered and had to be corrected afterwards. For messages from outside, the sender remains decisive, so that newsletters and distribution lists do not create wrong reporters.
• FixedThe watchers really are notified now. Participants once entered were neither saved nor informed — by the next time the ticket was opened the list was empty, although the journal showed the entry. Participants are now kept on the ticket: internal ones receive a notification in the application with every reply, external addresses receive customer-facing replies by e-mail. Internal notes stay internal.
• ImprovedUniform status colours and one status more. Status colours were stored separately in three places and contradicted one another: „Waiting for customer“ appeared differently in the list than in the ticket view, and the legend described a third thing again. It now applies throughout: blue = new, yellow = in progress, orange = waiting for customer or service provider, red = blocked, green = resolved, grey = closed. Newly added is „Blocked“ for tickets blocked internally; „Waiting for service provider“ can finally be filtered. In every waiting state the SLA clock pauses.
• ImprovedOut-of-office replies recognised more reliably. Automatic replies create neither a ticket nor a reply of their own. They are now recognised even when the sending system sets no technical marking at all. Discarded messages are logged with sender and subject, so that a misidentification stays findable.

2026.07.21.200621.07.2026

• NewWorkstations update themselves. A client installed on a PC of its own was not covered by the server update until now — not even when the update had been triggered from that very client. It now updates itself: a signature-checked manifest, every file secured individually by checksum, a restore point before the swap and an automatic return to the previous version if anything goes wrong. The files come through your own server — workstations need no internet access.
• FixedAn outdated workstation went unnoticed. The update check reported the version of the server, not that of the client. If the server was already running the new version, a workstation client left behind said nothing — even across several versions. It now compares its own version and points that out explicitly.

2026.07.21.191421.07.2026

• FixedSupport groups now show WHO is in them. Until now the administration listed only the users and roles recorded — but a role is a rule and not a person. If a group consisted of roles alone, it reported „4 members“ without showing a single name. The roles are now resolved to the people who hold them: for each person it says why they are in the group (directly, or through which role). Expired role assignments do not count towards it.
• ImprovedThe support groups page laid out afresh. The group overview at the top, below it the members of the group clicked — separated into the actual people and the membership rules, which can be removed. Adding and removing happens directly on the page; the previous dialog is gone. The OLA key figures remain beside it unchanged.

2026.07.21.183221.07.2026

• NewSeveral devices affected per ticket. A ticket still carries one main device — the SLA assignment, the analyses and the outage analysis hang on it. On top of that, any number of further configuration items affected can now be linked, in order to document the actual extent of an incident.
• NewPasting images and files straight into the journal. Screenshots, PDF files and text files can be pasted into the journal input field with Ctrl+V or dropped onto it. They are saved as an attachment to the ticket and appear in the list of attachments.
• NewMarking DHCP ranges within the subnet. Addresses lying within the DHCP range recorded for their subnet are marked as such in the address list. The marking is calculated and changes no data; it can be switched off for the whole tenant under Settings → IPAM and is active by default.
• ImprovedTidier ticket list. The administrative objects of the service desk — auto-routing, support groups, maintenance windows, approval levels, text blocks, the knowledge base and reporting — now sit together in the menu „Administration“ instead of as seven individual buttons. Data quality has moved into the ITSM dashboard, where it belongs as an analysis.
• ImprovedA saved view stays the default. A view once chosen is applied again automatically the next time the ticket list is opened — until another one is chosen.
• ImprovedThe members of a support group at a glance. The group overview now shows the number of members, the button carries a label, and changes that fail are reported in the dialog instead of going unnoticed.
• FixedThe tenant was invisible after the initial setup. A tenant created through the first-start wizard did not appear in the tenant administration, and issuing an integration token (for Ordivis Toolbox, for instance) failed with „unknown tenant“ — although the tenant name was displayed correctly. The tenant is now entered in the central catalogue immediately.
• FixedThe sorting of IP addresses. When sorting by IP address the addresses were compared character by character, which put 10.0.0.150 ahead of 10.0.0.2, for instance. Sorting now goes by the actual value of the address.
• FixedA reply from the reporter changes the status. If a reporter replied by e-mail, the message was only written into the journal. The ticket now returns from the waiting states and from „Resolved“ and „Closed“ back into progress; the SLA clock starts running again accordingly.
• FixedA missing label in the licence overview. In the licence detail view an internal placeholder appeared instead of the column heading.

2026.07.20.184920.07.2026

• NewManaging e-mail signatures centrally. Outgoing ticket mails now receive a uniform signature — globally and, if you wish, per support channel. Dynamic placeholders insert the agent (first name, shortened surname, function), the ticket number and the company details automatically; a legally complete imprint is appended to every mail. Administration and preview are found under Service Desk → E-mail signature.
• NewKeeping BSI Grundschutz++ up to date automatically. The security catalogue is reconciled daily with the official BSI source. If something changes, Ordivis Platform informs those responsible for the catalogue; the adoption deliberately happens only on confirmation (no silent switch). Your own adjustments to the catalogue are retained in the process.
• NewTechnical and organisational measures on the processing activity. For every processing activity the technical and organisational measures (TOMs) taken can be linked from a catalogue and set out in the record of processing activities under Art. 30(1)(g) GDPR.
• ImprovedFurther fields of use in the service desk. The ticket categories now additionally cover sales, facility management and human resources, with matching sub-categories — in existing tenants too.
• ImprovedFirst name and surname in the user profile. First name, surname and function or title can now be maintained directly in the user profile; among other things they feed the new e-mail signature.
• FixedThe first-start wizard despite demo tenants. If the demo tenants were installed along with the installation, the system wrongly counted as set up and the first-start wizard for your first own tenant, SuperAdmin included, failed to appear. Demo tenants are now left out of that check.
• FixedDeleting tenants reliably. Removing a tenant could fail or leave a half-deleted state behind. The operation now aborts in a controlled way if the database cannot be removed, and otherwise releases the database reliably.

Week 29 · 13–19 July 202645 versions

2026.07.19.195819.07.2026

• NewAsset tag scheme: putting the CI number together yourself. In the CMDB it can now be laid down per tenant how the CI number is built up — from building blocks that can be combined (fixed text, tenant, site, department, CI class, date and a running number with a selectable counter context). A configurator with a live preview shows the result immediately; if no scheme is active, the previous consecutive number stays. Changes apply only to CIs newly created — numbers already assigned remain unchanged.
• NewLabel template: setting the barcode type, the layout and the content. The appearance of the asset labels can now be configured per tenant: the barcode type (QR code, DataMatrix or Code 128), the layout template (small, standard or server tag), which fields appear (name, serial number, CI class) and whether the tenant's logo is printed on it. A PDF preview shows the label before printing.
• NewPrinting a label directly at the device. The CI detail view now has „print label" — a single asset label is generated according to the template and opened.

2026.07.19.154619.07.2026

• NewBackup: directory, schedule and retention can be set. The backup settings now show the directory actually in use and let you set the schedule (daily/weekly/monthly), the time and the retention period. The list of existing backups shows the date and the size correctly and can be sorted.
• NewLogs: where they are stored and how long they are kept. Administrators can lay down where the log files of the server services are written (to a data drive instead of C:, for instance) and how many generations of files, or what file size, are kept. The setting survives product updates and takes effect after a restart of the services.
• NewRestoring directly from the interface. A backup can now be restored directly from the settings. Because the whole database is replaced in the process, a permission of its own is needed — by default only for the SuperAdmin, but it can be granted to further people through the role and permission matrix.
• FixedThe backup path displayed, and empty table columns. The field for the backup directory showed a fixed default value instead of the path actually in use, and in the list of backups the name and the size stayed empty. Both are now displayed correctly.

2026.07.19.142219.07.2026

• ImprovedClearer start dashboard. The start page now brings the central key figures from every module together in one place – besides the CMDB and the service desk also IP management, information security/ISMS, emergency management (BCM) and data protection. The separate service desk chart tab is gone.
• ImprovedAnalysis and dashboard brought together. In the service desk there were „Reporting" and „Dashboard" twice over. The menu item „Dashboard" now opens the detailed analysis directly (SLA, handling times, satisfaction, backlog, agent load); the duplicate menu item is gone.
• ImprovedMore readable data protection impact assessment. In the DPIA tab the sliders for the residual risk now carry clear labels and no longer overlap; the field for the reasoning is considerably larger.
• FixedThe company name and the branding are saved again. Changes to the company name in the settings were, in some circumstances, not adopted. They are now saved reliably.
• FixedHits in the global search open the object. A click on a ticket, a CI or a user in the global search now leads straight to the object in question.
• FixedJournal entries in the portal are displayed with their content. In the ticket view of the self-service portal the journal entries stayed empty of content; their text is now displayed correctly.
• FixedUser search regardless of spelling. When assigning something – a device to a member of staff, for instance – the user is now found regardless of upper and lower case, by login name, display name or e-mail.
• FixedCI icons after signing in again. After signing out and in again without restarting the client, the icons of the configuration items could disappear; the class catalogue is now loaded afresh for each session.
• ImprovedAn icon for integration tokens. The entry „Integration tokens" in the platform settings now has a matching icon.

2026.07.19.121719.07.2026

• NewData protection: the processing systems on the activity. For every processing activity the IT systems doing the processing can be coupled in from the CMDB; their protection requirement is taken over from the information security module. If data especially worthy of protection is processed on a system without a sufficient protection requirement, Ordivis Platform warns of the „unprotected data flow".
• NewData protection impact assessment (DPIA). A guided threshold analysis under Art. 35 GDPR with a risk traffic light shows when an impact assessment is needed. An activity that has not been assessed deliberately stays grey — the traffic light never wrongly suggests „no concerns".
• NewMulti-stage approval, signed electronically. Depending on the target group (SME, enterprise, public administration), activities go through a single-stage or multi-stage approval workflow; every decision is signed electronically and recorded in an audit-proof way.
• NewSelf-service „My activities" in the portal. In the self-service portal, departments see the state of the processing they have reported — under review, approved or still incomplete — without any view of other people's activities.
• ImprovedA situation picture for the DPO, and target group profiles. The data protection centre shows key figure tiles (under review, overdue, special categories, third-country gaps), and every activity makes plain what the chosen target group profile provides for in terms of scope and approval stages.
• ImprovedA ticket status filter with multiple selection. In the ticket list several statuses can now be filtered at once („everything except closed", for instance) instead of just a single one.
• ImprovedDatabase performance on the hot paths. Additional indexes and a more aggressive clean-up (autovacuum) on the tables written to most often keep the ticket search fast even with large volumes of data; a representative capacity test measures the query shapes that really occur.

2026.07.18.232618.07.2026

• NewA capacity test for administrators. In the platform monitoring a SuperAdmin can now start a controlled capacity test against a throwaway database (never against real tenant data) and measure throughput, response times and the server's connection reserve — useful for demonstrating the sizing on your own hardware. The run is safeguarded (confirmation, fixed upper limits, a time limit) and is recommended outside business hours.
• NewThe most expensive database queries in the platform monitoring. The SuperAdmin monitoring now shows the most demanding database queries with their throughput and response time — a basis for finding performance bottlenecks in a targeted way.
• ImprovedDatabase performance and connection pools. The connection pool is now limited per tenant and runaway queries are capped by a runtime limit (database migrations are exempt from it). The operations manual describes the recommended settings for each size of customer.
• ImprovedThe foundation for further languages. The texts of notifications, e-mails and the ticket journal are now generated from language resources instead of sitting fixed in the program code — the technical foundation for extending Ordivis Platform by further languages in future. German and English stay unchanged.

2026.07.18.164118.07.2026

• FixedA more stable IPAM while scanning. When starting a network scan in IP management the client could crash in rare cases; the delayed reconciliation of results is now safeguarded. If loading the IP list of a subnet fails, a clear message now appears as well, instead of a list that is empty for no visible reason.

2026.07.18.152618.07.2026

• NewTaking network scan results from external tools into the IPAM. Ordivis Platform can now receive scan results from external administration tools – the free Ordivis Toolbox, for instance – through an interface. Remote subnets that the server itself cannot reach are created or added to in the process; devices found appear as unconfirmed entries and are reviewed and confirmed by an administrator. Sending works online and offline alike (with a buffer in the tool).
• SecurityRevocable integration tokens for machine access. For the connection a SuperAdmin issues an access under „Settings → Platform → Integration tokens" (no user login needed). It is valid for exactly one tenant, may feed in scan results and nothing else, can be revoked at any time and is stored only as a hash. Data fed in is not adopted blindly but confirmed by an administrator.

2026.07.18.124718.07.2026

• FixedUploaded attachments are shown in the list again – and can be deleted. In tickets and BCM emergency documents, uploaded files did not appear in the list of attachments (the upload itself worked). The cause was a matching error over upper and lower case. Attachments are listed again now and can be viewed, downloaded and – with edit permission – deleted.
• ImprovedIcons for the new platform settings. The entries „Platform monitoring", „HTTPS / transport security" and „API connection" now have matching icons in the settings navigation.

2026.07.18.115218.07.2026

• NewOperations manual and emergency manual. Ordivis Platform creates an IT operations manual from your modules automatically and maintains recovery plans that are handed straight to the checklist engine. Manuals can be exported as a branded PDF, go through an approval workflow and are reachable through an interface of their own in the client.
• NewHelp directly in the client. A context-sensitive in-app help (F1) with search, a PDF view and an about and support area – matching the version installed and usable without an internet connection.
• NewPlatform monitoring for SuperAdmins. An overview of the health of the system and its services, plus configurable thresholds and alerts for running the platform.
• SecurityAutomatic HTTPS certificates (Let's Encrypt) and HTTPS settings. Besides a certificate of your own and a reverse proxy, the API can now obtain and renew its certificate automatically through Let's Encrypt. The SuperAdmin sets the mode (reverse proxy / own certificate / Let's Encrypt) together with the domain conveniently in the interface; the installer asks for it during setup as well.
• NewThe client connection can be set freely. The address of the API server (host, port, HTTP/HTTPS) can be set during installation and at any time in the client – on the sign-in page and, as a SuperAdmin, under the settings, „Test connection" included. For internal networks with a self-signed certificate there is an additional option to ignore certificate errors deliberately (with a clear warning). That makes the client usable across sites over a VPN as well.
• NewAttaching files by drag & drop. Everywhere files are uploaded (tickets, the CMDB, process and emergency documents among others), several files can now be dropped at once by drag & drop – e-mails and attachments straight out of Outlook included.

2026.07.17.220217.07.2026

• SecurityHTTPS can be set on the service. The API and the portal can now either leave the TLS encryption to a reverse proxy in front of them (the default) or encrypt themselves with a certificate of their own. If a service runs unencrypted, a clear message points to the security recommendation. Obtaining certificates automatically through Let's Encrypt will follow in one of the next versions.
• FixedLess noise in the Windows event log. Several recurring warnings (the activation anchor, the HTTPS redirect, internal database notes) have been turned off – nothing changes in operation as a result.

2026.07.17.201817.07.2026

• ImprovedThe process model (BPMN) in a uniform light and dark design. The graphical process display (Processes → Process model) now follows the light or dark appearance of the application instead of a fixed white surface and thereby fits into the Material Design 3 foundation. Shapes, connections and labels are given clearly visible colours according to the mode. In the process editor the toolbar and the action buttons on each object are clearly recognisable in dark mode as well – previously they were barely legible white boxes. Nothing changes in the range of functions.

2026.07.17.190017.07.2026

• ImprovedA maintenance release – no functional change for your installation. This update concerns our delivery route and the demonstration data supplied with the product, and nothing else. The publishing process now shows us every check individually (manifest, signature, public reachability, the switch-over), and the signature check of the update manifest has been moved out into a lean component of its own – nothing changes in the program delivered as a result.
• ImprovedMore meaningful demo environment. The three demonstration tenants now show a usage history spanning several years – annual emergency exercises and Grundschutz checks, a statement of applicability (SoA) covering all Annex A controls, versions of the process models and checklists completed over the years. Visible only with the demo data switched on; productive installations do not contain it.

2026.07.17.171717.07.2026

• NewA crisis that has ended now leaves evidence behind. In emergency management the crisis dashboard shows situations that have been closed: when they began and ended, how long they lasted, the highest alert level they reached and who ended them. Until now, ending crisis mode deleted these details – of all things the basis for the follow-up under BSI 200-4 (review, lessons learned). The highest level reached is retained even if the situation eased again before the crisis team was stood down.
• FixedA data breach recorded but not yet assessed counted as not notifiable. As long as nobody had estimated the risk of a data breach, Ordivis Platform treated it as free of risk: the 72-hour deadline never became overdue, and the breach could be closed. Of all cases, „recorded and forgotten“ – the very case the deadline exists for – raised no warning. The duty to notify now depends on whether the risk was actually assessed – a breach that has not been assessed becomes overdue after 72 hours and cannot be closed quietly.
• FixedAfter a restart the publishing service forgot the state it had verified. Concerns our delivery route only, not your installation: after a restart the service reset its verification note and would have carried a release already fully verified as unverified again. It now reads the verified state back in at start-up.

2026.07.17.132917.07.2026

• FixedThe legal basis appeared as a technical abbreviation instead of in plain words. The activity profile showed „BerechtigtesInteresse“, for instance – while the export for the authorities correctly wrote „Art. 6(1)(f) (legitimate interest)“ in the same place. Both now show the same plain wording with the letter from the act, so that the assignment can be checked without a translation table.
• FixedThe internal publication check inspected the wrong release. Concerns our delivery route only, not your installation: the automatic check after uploading determined the release to be checked from the version file – but that file is deliberately switched over last. It therefore checked the previous release, long since complete, and reported „in order“ while the new one stayed unchecked. The check now follows what was actually uploaded.

2026.07.17.125417.07.2026

• FixedThe client crashed when opening an activity in the record of processing activities. A selection field in the activity profile (the target group profile) worked itself up endlessly against the display as soon as a stored profile was loaded – and the client shut down in the process without a message. It only became visible with real data: as long as no activity existed, none was ever opened.
• FixedThe same cause of crashes defused in twelve further places. Affected were selection fields in the CMDB (detail and list), the IPAM, the knowledge base, the role builder, the statement of applicability (SoA), the settings and the user list. None of them had come to light so far – they could have struck in day-to-day operation as soon as enough data was present. An automatic check now keeps the ground clear for good: it can no longer be built green if such a field is added unprotected.
• ImprovedFewer unnecessary queries while filtering. As a side effect of the correction, lists (the IPAM, the knowledge base, users) no longer reload when a filter is set to the same value as before.

2026.07.17.122317.07.2026

• FixedThe overview of the record of processing activities reported „complete“ although recipients were missing. The details about recipients from the previous update were checked in the activity itself and in the export – but the Overview went on computing their completeness from the legal basis, the categories and the deadline alone. Of all things, the working list of the data protection officer thereby sounded the all-clear where a mandatory detail was missing. It now reports „no recipients (Art. 30 I d)“ and „third country without safeguards (Chap. V)“ like every other view.

2026.07.17.115417.07.2026

• FixedThe record of processing activities was incomplete – and did not say so. Article 30 requires you to record the recipients of your data as well. That field did not exist until now, and worse: the findings column did not object to its absence. A record could therefore stand there „without findings“ and still be missing a mandatory detail – a certainty that a spreadsheet would not have feigned. We have fixed it and apologise for it.
• NewRecipients and third countries (Art. 30(1)(d)/(e)). For every activity you now record who receives the data – as a category, as the law requires („tax adviser“); you are free to add the specific name. Along with it the role, because the role decides the obligations: a processor (Art. 28) acts on instructions, a third party acting on its own responsibility needs a legal basis of its own – and a sister company is, in data protection law, a third party as well. „We pass nothing on“ is also a statement and wants recording; an empty field, by contrast, is not one.
• NewA warning for transfers abroad. If a recipient sits outside the EU and the EEA, the regulation demands a safeguard – an adequacy decision, standard contractual clauses, binding corporate rules or a derogation in the individual case. If it is missing, Ordivis Platform reports that as a finding. For the United States a reference to the adequacy decision alone is not enough: it does not apply to the country but only to certified recipients – which is why Ordivis Platform asks for the proof.
• ImprovedOrdivis Platform does not contradict your data protection officer. Whether a country has an adequacy decision is decided by the European Commission – and it revokes decisions again as well. A list of countries stored in the program might be out of date from the day it ships, and an out-of-date list that quietly says „permitted“ would be more dangerous than none at all. Automatically, therefore, we decide only whether a country belongs to the EU or the EEA. If you enter a decision for a country our edition does not know, you receive a note carrying a date – not a contradiction. The decision stays with you.
• ImprovedSample records in the demo tenants. Anyone demonstrating or trying out Ordivis Platform now finds worked-out examples in the data protection module instead of an empty list – payroll, applicant management, video surveillance, the residents' register, the allocation of nursery places and others, matching the target group profile in question. Deliberately among them: activities with open findings, so that it becomes visible how the module reports gaps.

2026.07.17.093317.07.2026

• NewData protection: the record of processing activities (Art. 30 GDPR). Anyone processing personal data has to keep a record of it – the supervisory authority may ask for it at any time. Ordivis Platform now keeps it in the product instead of in a spreadsheet: with the purpose, the legal basis under Art. 6, the categories of data subjects and of data, and the retention periods. Every change is verifiable field by field – who, when, which field, from which value to which. The evidence even survives the deletion of the activity: removing a processing operation does not wipe out the proof that it existed. Only someone with the permission of their own for it may approve – whoever records something does not approve it themselves.
• NewIncomplete activities are shown instead of coming to light during an inspection. For each activity the overview shows what it is missing for Art. 30: no legal basis, no categories, no retention period – and legitimate interest without the balancing of interests it requires. A filter turns that into the working list of the data protection officer. The special categories under Art. 9 (health, religion, trade union, biometrics …) and the criminal-law data under Art. 10 are supplied as a statutory catalogue – they are the text of the law, not configuration.
• NewNotification of personal data breaches (Art. 33/34). The first thing the list shows is the time remaining – because Art. 33 gives you 72 hours, and they run from the moment of becoming known, not from the incident. Both moments are recorded separately. „Not notifiable“ calls for a reason (Art. 33(5)), and a case cannot be closed while an obligation is still open: notifiable and not notified, or a high risk without the data subjects having been informed.
• NewAn export fit for the authorities at the push of a button. The record can be output as PDF, CSV or JSON (Art. 30(4)). The export reports the gaps along too instead of keeping quiet about them – the authority will find them anyway, only later and less well explained.
• NewDepartments report their processing themselves. A data protection officer cannot know every processing operation in a house on their own. In the portal a wizard leads the departments through the report in five steps – in everyday language, without prior knowledge of data protection („whose data do you process?“ instead of „categories of data subjects“). The report goes to the data protection officer as a draft for review.
• ImprovedNavigation sections fold up as soon as another one is opened. In the client all sections stayed open at once and the sidebar became hard to survey. It now behaves as in the portal: one section is open, the rest close.
• ImprovedThe internal message bus replaced. Invisible to you, but important: the library used until now will receive no more security patches from the end of 2026. It has been replaced by a maintained alternative – in good time and without any change to your data or your procedures.

2026.07.16.173116.07.2026

• NewSetting up two-factor sign-in yourself. Until now two-factor sign-in (MFA) could not be switched on for your own account at all – there was simply no route to it. Now you set it up yourself: in the account menu at the top right under Two-factor authentication. The dialog shows a QR code for your authenticator app and, beside it, the secret in plain text – anyone who cannot or will not scan types it in. Only the confirmed code from the app arms the second factor; an attempt broken off leaves your account unchanged and does not lock you out. Through the same menu you switch it off again as well (still only with a password or a valid code).
• FixedThe platform monitoring showed nothing at all any more. The dashboard stayed empty and reported an internal error. The cause: to determine the size of the database, every database on the server was measured – other people's included. On servers with many databases that took too long, ran into a timeout and dragged all the remaining key figures down with it, although those had long been available. Only what belongs to Ordivis Platform is measured now, the result is held for a few minutes – and if a single key figure cannot be determined, the dashboard shows a dash for it and delivers the rest all the same, instead of staying empty.
• ImprovedThe database is tuned during installation. If the installer brings PostgreSQL with it, it now sets the server up for operation right away instead of leaving it on the values it shipped with. The values are calculated from the target machine – memory, processor cores, free disk space – so that a large server and a small VM each receive settings that suit them. If you bring your own PostgreSQL instance or use an external database, its configuration stays untouched.
• FixedCleaning up old update versions left empty folders behind. When superseded update versions were removed the files were deleted, but the empty folders stayed lying on the server and accumulated. They are now cleared away as well.

2026.07.16.122616.07.2026

• FixedUpdates: services were not reliably waited for during the exchange. Windows reports a service that is in the middle of shutting down as „stopping“ – which counted as „stopped“ until now. The exchange of files could therefore begin while the service still had its files open; the update failed and was rolled back. Now the actual stop is waited for, a service is killed deliberately only if it does not stop by itself, and in every case the real end of the process is waited for before any files are touched. Product updates therefore run through noticeably more reliably.
• ImprovedA uniform icon in the taskbar and the task manager. Until now only the client and individual tools carried the Ordivis icon – the API, the portal, the update service and the progress window for updates appeared with the generic Windows placeholder. All the programs supplied now show the Ordivis icon throughout, in the title bar of the update window as well.

2026.07.16.101116.07.2026

• FixedSecurity: two-factor sign-in could be switched off too easily. Anyone who had activated two-factor sign-in (MFA) could until now remove it again with nothing but an existing session. Switching it off now calls for a fresh confirmation – the current password or a valid code from the authenticator app. A hijacked session can therefore no longer defeat the second factor unnoticed.
• NewAn IT emergency manual at the push of a button. Out of the emergency preparedness data already maintained for an information domain, a complete, versioned IT emergency manual under BSI 200-4 now arises: the emergency organisation (crisis team and emergency team with their members and deputies), the alerting and escalation chain, immediate measures, the communication plan, the contact directory and the applicable emergency documents. The decisive point: nothing is copied – the chapters refer to their source and show the current state every time they are called up. Change a telephone number in the alerting list and it is in the manual at once, without generating the manual again. No maintaining anything twice, no out-of-date emergency contacts.
• NewOperations and emergency manuals as versioned documents. Manuals can be kept as a draft and as an approved version: an approved manual is protected against changes, and a new version arises automatically for a revision. The basis for inspections and audits (BSI/ISO) in which the approved version has to be demonstrable.
• NewHelp directly in the application. The manual is now available directly in the client without an internet connection – built from the same sources as the printed manuals, and therefore always at the same level as the product.
• NewPlatform monitoring for operators. A new dashboard shows the state of the platform at a glance: database key figures (connections, cache hit ratio, long queries, size), active sessions and sign-ins, and per tenant the database size and the sessions. Refreshes itself automatically.
• NewSecurity events in the Windows event log. Ordivis Platform additionally writes security-relevant events (sign-ins, failed sign-ins, changes to passwords and permissions, service, update and licence events) into the Windows event log under a source of its own, „Ordivis Platform“. That allows them to be connected to an existing SIEM without extra effort; which events are written is configurable. The existing log inside the application remains unchanged.
• ImprovedThe reporter of a ticket can be corrected. The reporting contact can be set or changed afterwards if it was not known at first. For tickets arising from e-mails the reply address is evaluated as well, so that the actual customer is recognised as the reporter and not the mailbox doing the forwarding.

2026.07.15.225615.07.2026

• NewA visible progress window for product updates: An update triggered by hand now shows its course in a window of its own – with phases (backup, download, install, restart & check, completion), a progress bar and a live log – and reports when the services are available again. The progress stays visible even while the application itself is being exchanged and restarted; after a successful update the client comes back up automatically.
• FixedIn certain constellations the client was restarted invisibly in the background after an update — it now comes back up visibly and reliably. In addition the settings page shows the result of the last update correctly again.

2026.07.15.211915.07.2026

• NewA two-role licence model (admins & service desk agents): Named users are now licensed in two roles. An administrator licence covers the full range of functions including the service desk; a – new – service desk agent licence entitles the holder to work on tickets and the service desk only and grants no administrative access. Both roles are checked against the licence separately (with a clear message when the respective quota is reached); end users in the self-service portal remain unlimited and free of charge. Existing licences remain valid unchanged — a separate limit on agents only takes effect with a newly issued licence.

2026.07.15.173815.07.2026

• ImprovedMore robust product updates: An update is now carried out entirely by a standalone updater instead of partly by the running application server itself. Where needed the updater stops the application server service and the portal service, exchanges the files with an automatic fallback (rollback), checks every service for proper function after the restart and starts the client again last. That prevents incomplete partial updates (in which only the portal, but not the application server, had been updated, for instance) and shows the verified result per component in the settings.

2026.07.15.164415.07.2026

• FixedAn update stopped the application server from starting: In 2026.07.15.1549 a database migration inadvertently contained schema changes belonging to another module; on existing installations the start of the application server aborted after it had been applied. This build corrects the migration and contains the same range of functions as 2026.07.15.1549 (the checklist engine as well as the fix for notifications shown twice).

2026.07.15.154915.07.2026

• NewChecklist engine: Reusable checklist templates (sections, mandatory and optional items, responsibilities) can be assembled in the client, switched active and worked through step by step — with an audit-proof log (who, when, note) and completion by signature. Checklists assigned can be dealt with in the self-service portal; every run can be exported as PDF evidence. Typical scenarios (on-boarding and off-boarding, putting a server into service, recovery, change, emergency) are prepared as adaptable templates and are triggered automatically when a crisis is activated or a change is approved.
• FixedNotifications were shown twice: Messages that are additionally delivered by e-mail („update available", for instance) appeared twice in the notification bell. The bell now shows each message only once.

2026.07.15.124715.07.2026

• ImprovedA maintenance build without functional changes: An internal update; the reworked restart logic for updates from 2026.07.15.1130/.1217 (a single client restart, and only after the API is reachable again) remains unchanged and is confirmed in live operation with this update.

2026.07.15.121715.07.2026

• FixedAn automatic update no longer opens a client that was closed: With the restart fix from 2026.07.15.1130, an automatic update running in the background would have reopened a client that had deliberately been closed beforehand. The client is now only restarted after an update if it was also running when the update began.

2026.07.15.113015.07.2026

• FixedThe client restarts only once after an update: After an update through the client the program used to restart twice — first immediately on the previous version, then one or two minutes later, unannounced, a second time on the new one. The cause was two restart routes independent of each other. The client now restarts exactly once, and only after the application server (API) has been updated and is reachable again as a service; the update window points this out.
• ImprovedA note on this update: For technical reasons the fix only takes effect from the next update onwards — this one is still applied by the previous version and may therefore restart twice one last time. From the following update the client restarts only once, as described.

2026.07.15.100615.07.2026

• SecurityTenant separation hardened: The active tenant is now derived from the signed sign-in token and nothing else; a header value that used to be evaluated in addition is gone. Switching between tenants runs through the sign-in, where it is checked. No signed-in user can therefore reach another tenant's data any more by setting a header value.
• SecurityGranting permissions can no longer exceed your own: A role or a single permission can only be assigned if the person doing so holds that permission themselves — so nobody can obtain higher permissions for themselves through the role administration. Roles of other tenants can no longer be listed; the predefined system roles remain visible.
• SecurityThe sign-in no longer gives away valid account names: An unknown login name is now checked for as long as a known one with the wrong password. Which accounts exist can therefore no longer be read off the response time.
• SecurityAdditional tenant filters, closed by default: In further modules (e-mail and notifications, workflow, the change log, emergency planning) the tenant filters are now „fail-closed“ — even if a future query were to forget the filter, other tenants' data stays invisible. Background processes such as sending e-mail carry on unchanged.
• FixedError pages no longer give away internal details: In the event of an unexpected server error only a general message with a correlation ID appears; the technical particulars are in the server log and nowhere else. In addition, the update check ties the manifest more strictly to the version requested.

2026.07.14.232814.07.2026

• NewEvery component writes a log of its own: From now on the application server, the portal and the client each keep a log file of their own — named after their source (api-YYYYMMDD.log, portal-…, client-…). The files rotate automatically (daily and when the size is exceeded) and are deleted by themselves after the retention period set; they cannot fill the disk unnoticed. Set through appsettings.json: on or off, the level (debug, information, warning, error), the target folder and the retention. Until now the application server wrote its log to the console and nowhere else — as a Windows service, that is, into the void.
• FixedThe crash log of the client was lost: It was written to C:\ProgramData where, depending on the installation, the signed-in user has no write permission at all — the attempt failed silently, and of all things it was the crash report that was missing. On a terminal server, moreover, every user wrote into the same file. The log now sits in the user profile.
• ImprovedUpdate error messages say what really went wrong: Until now the client reported „no validly signed manifest“ on every abort — even when the manifest was perfectly fine and files were in truth missing on the update server. A message that claims something other than what the error is leads the search for the fault astray. Every abort now names its actual reason.

2026.07.14.213414.07.2026

• NewCopying the update log: If an update fails, a click on the log area in the update window is now enough — the whole text is thereby in the clipboard and can be pasted into a ticket or an e-mail. Until now it had to be typed out or photographed. This works for failed runs and successful ones alike.
• SecurityThe library for sign-in tokens updated: The component that checks the sign-in tokens (Microsoft.IdentityModel) was 19 minor versions behind and is now up to date.
• ImprovedOrdivis Platform transmits no error reports, crash data or system messages to the manufacturer — neither by itself nor at the push of a button. What leaves your house is decided by you alone, by copying the log and sending it yourself. The administration manual states this expressly.

2026.07.14.184514.07.2026

• SecurityUser administration secured: Creating users, assigning roles, resetting passwords and MFA and blocking sessions were protected by the sign-in alone — every signed-in user could thereby assign themselves the role „system administrator“. These calls now require the permissions „manage users“ and „manage roles“ respectively.
• SecurityAttachments: Who may read or delete an attachment now follows from the ticket it hangs on — no longer from a general module permission. Previously, attachments on other people's tickets could be retrieved and deleted, internal files invisible to the reporter included. Conversely, service desk agents can now reach the attachments on their own tickets, which was not possible until now.
• SecurityWrite permissions enforced: Change requests (CAB), process approvals, SLA profiles, routing and automation rules, approval levels, support groups, maintenance windows and text blocks could be changed with nothing but read permission. All calls that change something now require a matching write permission.
• SecuritySeparation of duties: A change request can no longer be approved by the applicant themselves (with the exception of emergency changes). For knowledge articles and process definitions the four-eyes principle already applied.
• SecurityIdentity from the sign-in: The tenant and the person acting are taken from the sign-in and from nothing else. Previously individual calls could supply both themselves — which made it possible, for instance, to attribute the decision in an access review to another person.
• SecurityThe signing key is checked: The service no longer starts with the placeholder key for sign-in tokens that ships with the product. For installations through the setup there is nothing to do — an individual key is generated there automatically (operations manual 3.1).
• SecurityAppointment and rating links: Acceptances and satisfaction ratings from e-mails are only saved after an explicit confirmation in the browser. Previously merely opening the link was enough — automatic e-mail scanning services could thereby accept or rate something unintentionally. The links are also limited in time.
• SecurityFile attachments checked: Executable files are rejected, the content is checked against the file extension, and the file type is determined on the server. Additional protective headers for the API and the portal.
• NewA new permission „manage service desk configuration“ (routing, automation, approval levels, groups, maintenance windows, text blocks) — for level 2 agents and administrators.
• NewNew permissions for the process documentation: „maintain processes“ (CMDB administrator) and „approve processes“ (CIO). Maintaining and approving are deliberately kept apart.
• NewLicences can be assigned to devices or users; the assignments feed into the compliance picture and appear in the 360° view of the asset.
• NewContracts can be edited in full in the detail view; when creating one, the term, the notice period, the annual costs and the cost centre can be recorded. Only then do the reminder for notice periods, the expiry warning and the cost analysis take effect.
• FixedWhen saving a configuration item, the purchase price, the purchase date, the cost centre, the start of the warranty and the support contract were lost — which made the depreciation and cost analyses useless.
• FixedContracts and licences lost their links when saved (owner, cost centre, contract reference).
• FixedExpiry warnings for contracts and warranties, and the decision on a change request, reached nobody — the notifications were created but not delivered. Warranty and contract warnings now remind in stages instead of daily.
• FixedThe bulk import of CIs and label printing demanded permissions that did not exist in the catalogue.
• FixedThe search in contracts, licences and suppliers distinguished upper and lower case („microsoft“ did not find „Microsoft“). Deleted contracts stayed visible in the list.
• FixedAmounts with a German decimal comma were falsified by a factor of 100 („1234,56“ was saved as 123456).
• ImprovedThree automatic checks secure the access control for good: no call that changes something without write permission, every permission demanded exists in the catalogue, and identity and tenant always come from the sign-in.

2026.07.14.144814.07.2026

• FixedIPAM: when creating a network or assigning an IP address, invalid entries (not a valid CIDR network, not a real IP address, an address outside the network) are now rejected – until now they were saved mutely and led to a wrong utilisation figure or to „no free IP“.
• FixedIPAM: an identical network can no longer be created twice within the same network area (VRF).
• FixedIPAM „next free IP“: DHCP reservations and the DHCP range are now skipped, so that no address out of the DHCP pool is handed out.
• ImprovedThe IPAM conflict check: it now recognises genuine double assignments (the same address on several CIs within the same network area) as well as addresses outside their network; the same address in separate network areas correctly no longer counts as a conflict.
• NewIPAM dashboard: the new tile „Conflicts“ shows the number of IP conflicts detected at a glance.
• ImprovedA more meaningful message when an IP address is already assigned or reserved.

2026.07.14.135214.07.2026

• FixedSwitching tenants as a platform administrator: the session now lasts – until now the client could show „no data / no permission“ after a while although the permissions and the data were there.
• FixedThe CMDB site tree: sites subordinate to a country or a region are displayed again (otherwise the tree stayed empty).
• FixedThe CMDB bulk action „set status“: works again; an inadmissible status change on individual CIs no longer aborts the whole action.
• ImprovedThe CI view: for things that are not devices (a country or region, processes, people, for instance) no device fields (serial number/manufacturer/model) and no network interfaces are shown any more.
• ImprovedThe new-CI wizard: the lifecycle status chosen is adopted, the name is asked for only once, and the buttons have been tidied up.
• NewThe network interfaces of a CI are given a speed with a unit (Kbit/s, Mbit/s, Gbit/s, Tbit/s).
• NewIPAM: a newly created subnet additionally appears as a CI and can thus be linked as a whole to a site, a router or a firewall.
• FixedIPAM: the DHCP marking on a single IP can be lifted again; „create CI from IP“ now also creates the corresponding interface and assigns the IP to it.
• ImprovedThe IPAM dashboard shows the utilisation per network; the utilisation marking in the prefix list is easier to read.
• ImprovedThe self-service portal: the navigation menu folds main categories up automatically when another one is opened.
• ImprovedAfter an update the client is only restarted once the API is reachable again (which prevents empty views directly after the update).

2026.07.14.115114.07.2026

• NewA set of automation rules in the service desk: configurable rules raise the priority automatically when a ticket is created, set the responsible group or agent, or add CC recipients – by category, subject or sender domain, for instance.
• NewFurther participants (watchers/CC) on tickets: internal colleagues and external addresses are notified along with everyone else on customer-facing activities.Addendum: Nobody was notified at the time. Participants once entered were not even saved — by the next time the ticket was opened the list was empty. Fixed in two steps: 2026.07.22.0844 (participants are kept, internal ones receive a notification in the application) and 2026.07.30.1933 (e-mail to internal participants as well).
• NewFulfilment tasks and checklists per service request – a ticket can only be closed once every task is done.
• NewPublic administration: a file or case reference and statutory deadlines (monitored separately from the SLA), a marking of the intake channel (De-Mail and beBPo among others) and a reference to the leading electronic file.
• NewSupplier SLAs (underpinning contracts) on the ticket with deadline monitoring of their own – including while the ticket is waiting for the service provider.
• NewTelephony: calls (incoming and outgoing, duration, note) can be logged directly on the ticket.
• ImprovedAutomatic satisfaction surveys (CSAT): after closing, the reporter receives a one-click survey by e-mail.
• ImprovedMore precise key figures: exact times of resolution and closure per ticket as well as a daily trend history (SLA attainment, throughput, backlog, CSAT) for trend charts.
• ImprovedFine-grained permissions per ticket action (assigning, approving and managing categories among others).
• ImprovedQuieter maintenance windows: during a planned maintenance window, automatic reminders for the tickets concerned are suppressed as well.

2026.07.14.092414.07.2026

• ImprovedFaster updates: update downloads now run in parallel and with transport compression – noticeably shorter update times, especially with many files. Security and integrity remain unchanged (a signed manifest, a checksum per file).

2026.07.14.084714.07.2026

• NewOver 35 new CI types: end devices (monitor, projector, scanner, thin client, IP telephone, UPS, video conferencing, backup system), enterprise objects (cloud subscription, cloud VM, SaaS service, cluster, Kubernetes, container, load balancer, storage volume, file share, WAN line, telephone system) as well as business and administrative objects (business service, IT service, SLA, OZG service, register, electronic file and filing plan, information domain, role, region, building services). The CMDB thereby covers small businesses, mid-sized companies, enterprises and public administration.
• ImprovedThe wizard for creating CIs now shows only the fields that suit the class – a room, a person or a process no longer gets a serial number field, for instance – and offers a target group filter (small business/SME/enterprise/public administration).
• ImprovedThe site hierarchy is stricter and more consistent: a floor, a room and a rack need a container above them, and assignments that do not fit (a floor inside a room, say) are prevented.
• ImprovedCI detail views now come uniformly from the class schema – no more deviating or permanently empty fields.
• FixedThe catalogue of selectable classes in the client matches the server exactly again: site and governance classes can be created, and abstract collective classes can no longer be chosen by mistake.

2026.07.13.220213.07.2026

• ImprovedThe portal navigation restructured: „My tickets“ now holds nothing but the actual actions; personal items (my devices, awareness, data protection) sit under „My workplace“, knowledge articles and known errors under „Knowledge & help“.
• ImprovedAnalyses (dashboards and reports) and specialist data views (the CMDB, topology, the change calendar, contracts, networks) are now separate; every compliance topic (ISMS, emergency management/BCM, disaster recovery, audit) is gathered under „Compliance & security“.
• ImprovedSettings (account, notifications, deputy) now sit together in one place.
• NewA notification bell with a counter of unread messages at the top right.

2026.07.13.212713.07.2026

• NewSatisfaction rating (CSAT): reporters rate a resolved ticket directly in the portal with stars and an optional comment.
• NewAppointments on the ticket: appointments proposed by the service desk can be accepted or declined in the portal and downloaded as a calendar invitation (.ics).
• New„My devices“: end users see the devices assigned to them (serial number and warranty included) and can report a fault directly as an incident.
• NewData protection self-service: requests for access to, rectification of or erasure of one's own data (GDPR Art. 15–17) are passed to the data protection officer.
• NewChange calendar: a read-only overview of planned changes and maintenance windows.
• NewKnowledge feedback: „Was this article helpful?“ with an optional comment right on the article.
• NewNotification settings: for each event, users decide for themselves whether they are informed in the portal and/or by e-mail.
• NewAwareness & acknowledgements: reading and acknowledging assigned policies and ISMS measures in the portal.
• NewApproval deputies: delegating approvals to a deputy for a period of absence.
• ImprovedThe portal can now be installed as an app (PWA) and keeps emergency documents available offline as well.

2026.07.13.162913.07.2026

• NewDocument branding: in the settings (Appearance › Branding) a company name and a logo can be stored per tenant. Both appear as a header on every generated document and PDF – report exports as well as ISMS and BCM reports (the risk register, protection requirements, the statement of applicability, the emergency plan and others). The footer continues to carry the Ordivis mark („Created by Ordivis Platform“) and the page number.
• ImprovedThe tenant administration is easier to survey: the tenant list and the health list as clear, scrollable tables with column headings and a status in plain words („Reachable“ instead of „True“, for instance).
• ImprovedUpdate settings: only the changelog scrolls now, while the version, the status and the options stay visible throughout.
• FixedUpdate notices could appear twice in some circumstances. The notification about a new version now happens reliably exactly once.

2026.07.13.151213.07.2026

• NewTenant health (SuperAdmin): a new dashboard in the tenant administration shows, per tenant, the database status, outstanding schema migrations and any quarantine. If the schema update of a tenant fails at start-up, only that one tenant is blocked (quarantine) – all the others carry on normally. „Repair“ allows the migration to be repeated for that tenant alone, without loss of data.
• ImprovedThe settings restructured: the long flat list of before is divided into five clear areas (platform, users & access, service desk, system & maintenance, appearance) – with icons and a search field for finding things quickly.
• Improved„Roles & permissions“ brings the role matrix and the object rules (ABAC) together under one item with two tabs.
• ImprovedThe authentication area has been slimmed from five tabs to three: AD/LDAP (including the synchronisation log), directory & role mapping, and MFA.
• ImprovedThe company name and the logo (branding) are now a section of their own, separate from the personal appearance settings.

2026.07.13.133113.07.2026

• NewObject rules (ABAC): fine-grained access control at object level. For each role and permission, rules can be defined as an expression („own objects only“, „amount ≤ 500“, for instance) that are checked in addition to the coarse permission. Managed in the client under Settings › Object rules, with a guided dialog for creating them and a live check of the expression.
• NewTenant hierarchy: tenants can be assigned to a tenant above them (a holding or parent company) – with protection against cycles. A roll-up overview aggregates key figures across the entire subtree.
• ImprovedDeleting a tenant safely: before the final deletion a GDPR export of the tenant database is created and kept automatically; an audit-proof audit entry arises in addition. If the export fails, nothing is deleted.
• ImprovedCross-tenant grants (holding/MSP): tenants with cross-tenant access active now appear in the tenant selection of the user who has been granted it.
• ImprovedBackground processing is tenant-aware throughout: fetching e-mail, sending it and processing events all run per tenant against that tenant's own database.
• ImprovedOperations: the release process shows a progress window with phases and, after the transfer, automatically checks that all files are complete (by size) together with a sample of hashes, guarding against transfer errors.

2026.07.13.110613.07.2026

• NewColour schemes: every user can choose between six coordinated schemes – blue, turquoise, indigo, neutral, warm and high contrast – each in a light and a dark variant. Switching takes effect at once (a live preview) and is saved across devices. In the client under Settings › Appearance, in the portal through the palette menu in the header.
• NewAccessibility: new personal options for high contrast (it picks up the Windows high contrast mode as well), an adjustable font size (80–150 %), reduced motion and a permanently visible keyboard focus.
• ImprovedIcon buttons without visible text now carry a name for screen readers – which makes the application considerably easier to operate with assistive technology.
• ImprovedEvery colour combination has been checked and secured against the contrast standard WCAG 2.1 AA; two existing weaknesses of contrast in the standard blue were fixed in the process.

2026.07.13.081913.07.2026

• NewModule administration: in the settings under „Modules“ the SuperAdmin can switch individual function modules (service desk, knowledge base, IPAM, discovery, contracts & assets, BCM, recovery planning, ISMS, workflow) on and off system-wide, or differently per tenant. Modules switched off disappear from the navigation of the client and the portal. Mandatory modules (CMDB, identity, audit, notifications, activation, licensing, reports) always stay active.
• ImprovedThe update window now appears in the dark Ordivis design with a title bar of its own in matching colours and the product icon – instead of the light system title bar.
• FixedIn the audit log the user acting was sometimes shown as an ID number instead of by name. The plain name now appears throughout – for older entries as well.
• ImprovedAll the tabs in the application have been given a uniform look – as in the process detail view.

Week 28 · 6–12 July 202629 versions

2026.07.12.224312.07.2026

• FixedThe „update available" notice was laid over the title bar and covered the logo, the quick actions and the search. It now appears in a bar of its own beneath the menu bar and overlays nothing any more.
• ImprovedThe notice appears only once (no stacking when the check runs repeatedly) and disappears reliably as soon as no update is outstanding.

2026.07.12.215712.07.2026

• NewVersion traffic light in the version information: a coloured bar plus a status in plain words – green (current), yellow (update available), orange (close to the minimum version, urgent), red (no longer supported).
• ImprovedSettings ▸ Updates now checks automatically when opened and loads the version, the components and the changelog together while doing so – no more switching pages.
• ImprovedThe changelog is given more room (two thirds of the width) and, with coloured category markers on each entry, is considerably easier to read.

2026.07.12.205712.07.2026

• NewRolling support window: version.json carries a minimum version; the client indicates in stages whether the version running is supported. „Ageing" warns early and predictably, „no longer supported" insists on the update firmly – to your benefit (security, bug fixes, features).
• SecuritySecurity releases are marked as mandatory and raise the supported minimum version immediately. The enforcement compels the update but never blocks day-to-day operation and is safe offline and in air-gapped environments.
• ImprovedAutomatic, reference-based clean-up logic for the delta store (it keeps the current version, the minimum version and the newest releases; it removes orphaned blobs). Customers are never locked out, because updates always aim at the complete newest manifest.

2026.07.12.193512.07.2026

• NewManual update as an option of its own: a click on „update manually now" checks for a new version at once and applies a validly signed delta directly – regardless of the auto-update behaviour that has been set.
• NewSeveral maintenance times per day: for the fully automatic update, any number of times can now be defined (02:00 and 14:00, for instance).
• FixedThe version currently installed is displayed correctly in the update centre again (instead of „unknown").
• FixedThe component overview now lists every component installed (the API, the web portal, the discovery worker, the client) with the version it is running – even when no update is outstanding – instead of the API alone.
• ImprovedThe changelog is shown as a column of its own beside the version, the auto-update policy and the components.

2026.07.12.181812.07.2026

• NewAn update centre in the client: the version available, the components affected (API, portal, worker, client) with their respective versions and the changelog prepared for reading – all in one place under Settings ▸ Updates.
• NewThree-level auto-update policy (off / notify only / fully automatic): with the fully automatic setting the server applies the signature-checked delta by itself in the chosen maintenance window (weekdays plus time of day) – with a pre-update backup, a health check and an automatic rollback.
• NewIntegrated changelog: alongside the version information, the website keeps a continuously updated changelog (changelog.json); the client displays it prepared for reading and highlights entries newer than the version installed.
• ImprovedThe update area replaces the non-functional placeholders of before (the channel selection, the auto-check switch) with real settings that are actually saved.

2026.07.12.173012.07.2026

• NewControlled delta self-update of the running installation, verified live: staging → a decoupled bootstrapper → a health check → rollback, without a manual full setup.
• FixedAn update progress window of its own in the Ordivis design; the Ctrl+F tooltip is suppressed.
• SecurityThe vault key is protected against deletion by an update; secrets sit in the machine environment instead of in configuration files.

2026.07.12.105912.07.2026

• NewISO/IEC 27001:2022 in full: a statement of applicability (SoA) covering all 93 Annex A controls, with an obligation to justify exclusions and a versioned export, the ISMS context (scope, interested parties, policy) per information domain, and a risk treatment plan linked to the Annex A controls.
• NewISMS management system (clauses 9–10 plus 6.2/7.5): measurable objectives and KPIs, an internal audit with findings, a management review with approval by the leadership, nonconformities with corrective action and a review of effectiveness, and document control.
• NewCertification readiness: a „ready for stage 1" traffic light with clause coverage and a display of gaps; PDF reports for the SoA, the risk treatment plan, the internal audit and the management review.
• ImprovedNotifications in the client can now be deleted – individually or all those already read, so that the bell list does not grow without end.
• FixedProduct licensing: the hardware fingerprint is now stable across reboots (physical network cards only) – the wrongful „tampering detected" block after a restart is gone; there is self-healing on top of that when hardware is legitimately changed.

2026.07.12.040012.07.2026

• NewComplete CI templates for every category: each CI class brings a worked-out set of attributes with it, oriented towards industry standards (ITIL SACM, ISO 19770/55000, BSI) – mobile device, storage, router, firewall, access point, VLAN, operating system, database, application or specialist procedure, web service or API, rack, data centre, department and manufacturer or supplier among them.
• NewSearchable CI picker for reference fields (owner, manufacturer, site, the person assigned): type a name and pick it instead of a technical ID – when creating, in the detail page and when reclassifying.
• ImprovedThroughout the client, tabs are uniformly recognisable as one coherent surface (processes, contracts, licences, settings, CI classes, dashboard).
• FixedSystem health: the e-mail check now tests the outgoing mailbox actually configured instead of an unused fallback port – „degraded" despite a correct mail setup is gone.

2026.07.12.020012.07.2026

• NewProcess management in full (business process management/lean/ISO 9001): the process detail page is divided into three tabs – profile, Process model (BPMN) and Documents.
• NewBuilt-in graphical BPMN editor: drawing processes by drag & drop (a swimlane template, a German interface), with a naming linter and internal-control checkpoints right on the element; saving as a draft with four-eyes approval. In the web portal as well (SVG and PNG export).
• NewThe profile as the central overview with all the metadata at a glance and directly editable; RACI with a searchable dropdown of people; a visible PDCA cycle with follow-ups to the process owners.
• NewPDF viewer: PDF documents open in a window of their own (in a new tab in the portal) instead of being downloaded.
• ImprovedTabs are everywhere easier to recognise as one coherent surface; the website menu gathers the product, the features and the roadmap under „Product".

2026.07.11.183411.07.2026

• NewOnline activation active by default: new installations register with the activation service and fetch a confirmation; the licence status now shows „Online activation: yes“. The check is advisory – without a network connection the installation remains fully valid through the licence file.
• FixedLicence status: the field „online activation" was always shown as inactive despite a valid confirmation; it now reflects the actual online state.
• NewAn installer option: during setup you can optionally create the three demo tenants (a small business, a corporate group, a municipal administration) with sample data.

2026.07.11.155511.07.2026

• ImprovedAn appointment with the reporter: the action „propose an appointment" is easier to find now – it sits in the new menu „Actions“ on the ticket (together with escalating and raising to a problem).
• FixedCreating a tenant: provisioning a database of its own per tenant now works reliably; the operations manual describes the database permission required.

2026.07.11.153811.07.2026

• NewPermission matrix: roles and their permissions are managed in a clear matrix (grouped by module). Roles of your own can be created, filled and deleted; system roles are write-protected.
• NewTenant administration & SuperAdmin: the first user of the installation (the SuperAdmin) creates, archives and deletes tenants through an interface of their own and can delegate these platform permissions to a platform administrator in a targeted way. Ordinary IT administrators deliberately do not receive them.
• NewTenant switcher: through the title bar the SuperAdmin moves between all tenants without a restart.
• NewGDPR data subject rights per tenant: structured access to data, anonymisation („the right to be forgotten") and the record of processing activities.
• ImprovedAccess control: an enforced object level (ABAC) as well as an automated check that no endpoint is reachable without a permission check. It also fixes a possible cross-tenant escalation of privileges.

2026.07.10.200710.07.2026

• NewProposing an appointment on the ticket: a service desk agent proposes an appointment to the reporter from within the ticket. The reporter receives a calendar invitation (.ics) and can accept or decline with a click – without signing in. The answer is noted on the ticket; if accepted, the appointment is registered for the agent as a follow-up.
• NewAutomatic update check: the application compares the version installed with the one published on the website. If a newer one is available, the administrators are notified (in the application, in the portal and by e-mail) – each channel can be set individually.

2026.07.10.192510.07.2026

• NewThe process profile now links everything that is already there: the systems used, the interfaces to other processes and the suppliers and customers (SIPOC) are picked through searchable selection fields — as are the process owner and the process manager, as real users. Only the link is ever saved, never a copy; names always appear live.

2026.07.10.190610.07.2026

• ImprovedA new, tidied-up navigation: business processes now have a central place of their own, „Processes" — creating them, the profile, the key figures and the graphical BPMN model in one place. The topology views have moved into the CMDB, the DR planning into emergency management, and the governance areas stand together.

2026.07.10.182010.07.2026

• NewBusiness processes are displayed graphically as a BPMN diagram. Embedded in it is the open-source industry standard bpmn-js (Camunda); it runs entirely in the client, without any connection to the internet.
• FixedSeveral labels in the client were displayed wrongly after translation: the navigation read „Assets & Lizenzen" instead of „Assets & Lizenzen", and in individual places an internal key (such as „L944d740e1b") appeared instead of the text. All the labels concerned now show correct German or English.

2026.07.10.173110.07.2026

• NewProcess models in BPMN format: a model from any tool is imported and versioned. An approved version stays unchanged – changes arise as a new version, and approval follows the four-eyes principle.
• NewA check of the naming conventions: a step is called „check invoice", an event „invoice checked", and a gateway is a question with labelled answers. The check points out deviations without blocking the work.
• NewAn internal control system on the process step: control points with a type (preventive or detective), a mechanism, a frequency, an owner, the evidence kept and the line of defence. A control without evidence is not accepted, because it would be worthless in an audit. Controls refer to the requirements of the security management instead of copying them.
• NewTarget versus actual: an event log from day-to-day operation is held up against the model. What becomes visible is what happens unplanned, what is modelled and never occurs, which path is the most frequent and at which step the process hangs the longest.

2026.07.10.165410.07.2026

• NewThe protection requirement is inherited along the dependencies: what a business process needs applies to the servers, applications and databases carrying it as well — across any number of levels. Before it is adopted, a preview shows which increase comes from where; nothing is ever lowered.
• NewThe IT-Grundschutz now knows the objects a classical CMDB does not keep either: the management system itself, information and data holdings together with their confidentiality level, controlled documents, certificates, cryptographic keys and digital identities.
• NewMachine-readable evidence in OSCAL format: the security concept with one component per object, the results of an audit and the action plan with owners and deadlines — retrievable as JSON, without the detour through reports.
• NewProfiles for the check: basic, core and standard protection are supplied with the product. Profiles of your own exclude individual requirements with a reason given; after a catalogue update a delta view shows which exclusions now point into the void.
• NewBusiness process documentation: a process profile including the ISO 9001 details, a RACI matrix with exactly one accountable role, a CRUD matrix with data ownership, key figures with a target and an actual value, and a follow-up that gives a reminder when reviews fall due. The impact analysis answers which processes fail if a particular system fails.

2026.07.10.154910.07.2026

• ImprovedThe self-service portal now points out the state of the licence: if the trial period is running out, a warning appears; if no valid licence is present, a clear notice that changes are blocked and that reading remains possible. Previously, entries failed in that state without a word.

2026.07.10.151210.07.2026

• NewStocktaking with QR labels: a sheet of labels (A4) is printed as a PDF, and the QR code carries the inventory number. During the stocktake the code is scanned – an ordinary hand scanner behaves like a keyboard, no special software is needed – or entered by hand. The discrepancy list shows continuously what has been recorded, what is missing and what turns up unexpectedly; closing the run with a note freezes it as a record. Devices that were found thereby count as inspected at the same time.
• NewKnowledge articles can be translated per language. The article itself remains the master version; if the reader's language is missing, the master version is delivered. The self-service portal picks the version automatically from the language that has been set.
• FixedIn the self-service portal older ticket texts were always rendered as Markdown. Characters such as * or # from the time before that were thereby misread as formatting. Such texts are now displayed unchanged.

2026.07.10.140610.07.2026

• NewISO/IEC 27001:2022 is usable out of the box: the complete Annex A catalogue with all 93 controls (organisational, people, physical, technological) as well as the management clauses 4–10 is created in every installation – with no import. The 38 technological controls are already linked to CMDB CI classes, so that target objects can be proposed automatically.
• NewMailbox connection to Gmail and Microsoft 365 via OAuth2 (XOAUTH2): the authorisation happens through a device code in the browser, and no plain-text password is stored anywhere. The refresh token sits encrypted in the vault, and the access token is renewed automatically.
• NewReliable e-mail sending through a queue: if the mail server fails briefly, no ticket reply is lost any more – delivery is retried at growing intervals. A sending limit per mailbox prevents the provider from blocking the account because of too many messages.
• NewNotification settings in the self-service portal: for each event, every user decides separately whether they receive a notification in the application and/or an e-mail. Events calling for a reaction outside the application (an SLA breach, an approval, crisis mode) are preset.
• NewProtection for the ticket mailbox: out-of-office replies, non-delivery reports and distribution lists no longer create a ticket, mail loops are prevented, and a single sender can no longer flood the mailbox. Attachments are checked for size and type before being saved; executable files are rejected and archives are only shown internally. A virus scanner can optionally be hooked in.
• NewMailbox monitoring: after three failed retrievals the administrators receive a message; after ten the mailbox is shut down instead of continuing to address a blocked account. Expiring OAuth authorisations are announced seven days in advance.
• FixedSecurity: The certificate setting „allow your own certification authority" in truth accepted every faulty server certificate – the fingerprint stored was never checked. It now really does check it and rejects a certificate issued to a different server in every case.
• FixedIf sending a ticket reply failed, the error was merely logged and the message was lost. It is now delivered again.
• FixedNotifications switched off in the application were delivered all the same: the user's setting was read but not heeded.
• ImprovedThe key for encrypting the mailbox credentials can now be configured explicitly and changed during operation. When it is changed, every secret is re-encrypted; if anything fails in the process, the previous key stays valid and nothing is altered.

2026.07.10.083010.07.2026

• NewEmergency management (BCM) in full: a business impact analysis with a rule-based pre-filter (time-critical or not), detection of time chains and SPoFs on the CMDB dependency graph, and the target-actual gap (the recovery time achievable versus the one required).
• NewA damage matrix (BSI 200-4) with an automatic MTPD proposal and RTO inheritance across the process hierarchy; open continuity gaps can be taken over as risks straight into the ISMS risk analysis.
• NewA full response organisation (crisis team, emergency team, response teams) with a deputy check (n+1) and a role for the NIS 2 duty to notify; document control with an approval workflow and interactive recovery playbooks that can be ticked off in an audit-proof way.
• NewExercise and test management (tabletop, alerting, staff framework exercise, full test) with lessons learned and CAPA actions, proof of conformity „at the push of a button" and a management summary as a PDF.
• NewMunicipal & critical infrastructure: a one-click administrative crisis team (sections S1–S6), crisis mode with alerting of the team members, and an audit-proof operations log (append-only) with a digital situation display.
• NewMultilingualism (DE/EN) in the client and the self-service portal, switchable per user; notifications and e-mails are delivered in the recipient's language. Further languages can be added without changing the program.
• ImprovedISMS reporting completed: a delta report (the progress between two Grundschutz checks) and a CI detail tab „Information security"; if a linked CI is decommissioned, the information security officer receives a note to review it. CMDB class templates can be duplicated and exported and imported as JSON.

2026.7.9.191009.07.2026

• NewAn operations dashboard in the portal: live key figures of the service desk (open, in progress, SLA breaches, major incidents, unassigned, changes awaiting approval) as well as analyses over 7/30/90 days with charts (backlog by age, the spread of CSAT, agent load) – governed by permissions.
• NewThe portal analyses extended: known errors (KEDB) with a workaround search, an interactive dependency and topology map (read-only), a network & discovery overview (subnet utilisation, the latest scans) and audit-proof research in the audit log – each governed by permissions and confined to your own tenant.
• NewThe complete ticket journal in the portal: the whole history (who, when, what) is displayed chronologically; internal notes remain visible to authorised agents alone.
• NewA Markdown editor for large text entries (ticket replies, descriptions, service requests) with a live preview; descriptions and the journal are displayed with their formatting.
• ImprovedPersonalisation & ease of use: the light or dark design chosen and the state of the menu are kept per user; „Print / PDF" for dashboards; accessibility (a skip-to-content link, a visible keyboard focus); the duplicate logo removed.
• FixedThe audit log is strictly confined to your own tenant (no cross-tenant access) and the history of deep network scans has been corrected (missing database tables are now created).

2026.7.9.173609.07.2026

• NewThe service catalogue is enterprise-ready with a guided ordering wizard: users pick a preconfigured service and are led step by step through a form specific to it (mandatory fields, selection lists, dates, yes/no). Out of that a correctly categorised service request ticket arises automatically in the service desk – including routing to the responsible group and forwarding for approval where that is recorded.
• NewPreconfigured catalogue services, public administration included: access to a specialist procedure, a staff ID card or smart card and time-limited, logged access for external service providers (VPN) among others – alongside classic IT requests such as a password reset, a software installation, a new laptop and onboarding.
• NewA portal operations dashboard: live key figures of the service desk (open, in progress, SLA breaches, major incidents, unassigned, changes awaiting approval) as well as analyses over 7/30/90 days with charts (backlog by age, the spread of CSAT, agent load) – governed by permissions.
• ImprovedSelf-service portal: collapsible menu sections, a new page „My account" for changing your own password (local accounts; AD/LDAP accounts get a note) and a CMDB search with wildcards (* and ?) – with no input at all, every CI is shown.
• FixedThroughout the portal, meaningful names are shown instead of technical IDs (the selection of an information domain in the ISMS dashboard and in reports, for instance).

2026.7.8.230408.07.2026

• NewISMS – reporting: four audit-ready PDF reports per information domain – a management summary, the risk register, the action plan (from the IT-Grundschutz check) and an overview of protection requirements (confidentiality, integrity and availability per target object).
• NewISMS dashboard: the spread of protection requirements, the progress of the Grundschutz check (checked, met, open measures) and the risk situation (acceptable, tolerable, critical) per information domain – at a glance as key figures and distribution bars.
• ImprovedBusiness processes in the CMDB: a guided wizard for creating them (master data → profile → classification) and a colour-coded process icon of its own (core = blue, support = green, management = orange).

2026.7.8.181408.07.2026

• NewISMS – risk analysis (BSI Standard 200-3): assessing risks per information domain (frequency of occurrence × extent of damage → a risk value of 1–25 and a category), recording the treatment strategy and the residual risk, and having risks approved by the information security officer (risk acceptance). If the residual risk is critical, a notification is triggered automatically.
• NewA graphical 5×5 risk heatmap – the risk profile (green/yellow/red) at a glance, switchable between gross and residual risk.
• NewA catalogue of threats supplied with the product: the 47 elementary BSI threats (G 0.1–G 0.47) as well as an enterprise threat catalogue of our own with 25 modern threats (the supply chain, APTs & AI-driven attacks, cloud and shared responsibility, regulation & geopolitics, insiders, plus scenarios for SMEs, public administration and critical infrastructure) – selectable directly in the risk dialog and extensible in the catalogue.
• NewSelf-service portal: approved emergency documents (emergency plans, playbooks, checklists) can be retrieved by authorised users – grouped by type of document, with a download; confidential documents stay hidden.

2026.7.8.131308.07.2026

• NewEmergency and continuity management (BCM) under BSI Standard 200-4 and ISO 22301: a new module with a role of its own, „BCM officer", and fine-grained permissions – the foundation stone for the digital administrative and crisis team.
• NewScope & process links: business processes are kept as CMDB objects and assigned to the BCM read-only – nothing maintained twice. Business processes are created through one shared procedure in the CMDB and the BCM.
• NewCriticality figures per process: MTPD, RTO, RPO and MBCO together with immediate measures – the basis for the business impact analysis and for setting priorities.
• NewContact & alerting lists with basic escalation – built on the contacts already there.
• NewA library of emergency documents with a basic emergency plan as a PDF, and an encrypted offline emergency kit (generated automatically by a scheduled background job) – the plans stay available even in a total outage.

2026.7.7.203007.07.2026

• NewInformation security (ISMS / BSI IT-Grundschutz) – information domain & protection requirements: taking business-critical objects straight out of the CMDB as target objects and determining the protection requirement (confidentiality, integrity, availability) by the maximum principle – with nothing maintained twice.
• NewISMS – modelling: assigning the matching modules of the BSI Grundschutz catalogue to the target objects, with an automatic proposal based on the CI class.
• NewISMS – the IT-Grundschutz check & action plan: a systematic target-actual review of every requirement (yes / partly / no / not applicable), an automatic action plan with status tracking and an audit-proof conclusion of the audit.
• ImprovedThe menu area „Information security" is now cleanly structured (security catalogue · information domain · Grundschutz check) – consistent with the other modules.

2026.7.7.142007.07.2026

• NewSystem status & announcements: a central status page (in the client and the self-service portal) shows the operating state of every service (API, database, e-mail gateway, message bus, scheduled tasks) in real time. Administrators can publish platform-wide announcements and maintenance windows – visible across tenants.
• NewA new start dashboard in the client with tabs and tiles – consolidated key figures including the service desk wallboard at a glance.
• NewEvery table can now be filtered and sorted throughout, with freely selectable columns and an export as CSV or PDF; columns can be moved with the mouse and adjusted in width.
• NewSelf-service portal: a personal dashboard (your own tickets, tasks, the CIs & assets assigned to you, expiring contracts) as well as creating tickets and a personal ticket overview.
• ImprovedThe navigation in the client tightened: ITSM and the service desk brought together under one menu, the quick actions moved into the header, and the current tenant shown in the title bar.
• ImprovedThe self-service portal now follows the same Material Design 3 colour scheme as the client, dark mode included.

Week 27 · 29 June – 5 July 20266 versions

2026.7.4.181604.07.2026

• FixedEnd users could not create tickets because of a permission check that was too strict (access denied). Creating one now correctly requires only the permission „create tickets".

2026.7.4.174904.07.2026

• NewIPAM: subnets can now be deleted in full – including every IP address and DHCP reservation they contain.
• ImprovedThe network scan now covers complete /16 networks and finds every reachable host (until now the scan was limited to a part of the range).
• ImprovedLists of IP addresses are sorted numerically and correctly (.2 before .16 before .150, for instance) instead of alphabetically.
• ImprovedSessions stay active for good: the access token is renewed transparently in the background – no more unexpected interruption when it expires.
• ImprovedOnly one active sign-in per user (single login); a new sign-in ends the previous one, and the session overview shows genuinely active sessions and nothing else.
• FixedAudit log: the details of a change are saved reliably again (certain entries used to be discarded).

2026.7.3.164103.07.2026

• NewInformation security module (ISMS): the import of the BSI IT-Grundschutz security catalogue (OSCAL) including its modules, requirements and the role of information security officer.
• ImprovedPermission gating (RBAC) extended across the modules to read access as well – a consistent 401/403 guard on every module endpoint.
• NewDemonstration data sets (small business, enterprise, municipal) for presentations and test installations.

2026.7.2.125202.07.2026

• ImprovedThe migration of the whole platform to .NET 10 (EF Core / Npgsql 10) – a current runtime with long-term maintenance.
• NewProduct activation extended: HA and cluster enforcement, re-activation after a change of hardware, licence telemetry.

2026.7.1.135401.07.2026

• NewReceiving e-mail (IMAP & POP3): incoming mails create tickets automatically, including routing by group and tenant.
• NewForgery-proof matching of replies through a signed reply token in the subject.
• NewGroup-based ticket visibility: agents see the tickets of their support group or groups.

2026.6.30.175730.06.2026

• NewAuthenticated deep discovery (SSH, SNMP, WMI/CIM, LDAP/AD) with an encrypted credentials vault.
• NewSoftware inventory for counting licences, including a CI detail tab „Installed software".
• NewInventory snapshots & diffs per CI (history and comparison).
• ImprovedAn audit log view and report administration (export as XLSX/CSV/PDF) in the client.

Week 26 · 22–28 June 20261 version

2026.6.2424.06.2026

• NewThe knowledge base in the self-service portal: public articles, search, browsing by category and deep links.
• NewProduct activation and trial (an offline MVP) with a signed licence file (ECDSA).

Week 24 · 8–14 June 20263 versions

2026.6.1313.06.2026

• FixedThe stability of the WinUI client at start-up (resources and design tokens) considerably improved.
• NewSelf-service: changing your password, and a password reset by an administrator (including unblocking).

2026.6.1212.06.2026

• NewA CI template engine with 26 templates (for SMEs & mid-sized companies) and a class editor.
• NewA workflow engine with escalation and triggers on CI attributes.
• NewVisualisation of dependencies and business services (a topology map).
• ImprovedMultilingualism (DE/EN); a clean installation on Windows Server 2025 verified.

2026.6.1010.06.2026

• NewThe foundation published: the CMDB (SACM), the service desk (incident/problem/change/service request), IPAM, asset, contract and licence management, DR planning.
• ImprovedSecurity hardening (an AES-256 vault, brute-force protection, HSTS, a gapless audit log).

Version scheme: YEAR.MONTH.DAY.TIME. Older maintenance builds without functional changes are summarised here. A machine-readable version feed: version.json.