Roadmaps describe what a project intends to build. They are cheap, and everyone has one. Rather more informative is a list of what a project has bound itself never to build — because every item on such a list is a revenue idea someone will one day propose in a board meeting, and the list is the pre-committed answer.
OrdinateDB's non-goals are part of the governance documents, not the marketing. Here is each one, with its reasoning, and — more usefully — the temptation it forecloses.
No silent lossy compression
The reasoning. A historian's entire value is that the archive can be trusted more than memory. Exception and compression deviation, applied silently, make the archive quietly less trustworthy than a paper chart recorder — we worked a concrete example in the case against silent compression. If data is ever thinned, it is through a lifecycle policy: visible, versioned, approved by a named identity, and disclosed in every affected query result.
The temptation foreclosed. "Storage-optimised tier — 90% smaller archives!" It is a genuinely sellable feature. It is also, precisely, the defect we exist to correct, sold back as an upgrade.
No tag-count licensing
The reasoning. Metering tags taxes the product's purpose. Every plant that pays per tag instruments less than the plant deserves; engineers ration series the way nobody rations columns in a database. The whole failure pattern — audits at renewal, deleting history to duck a licence band — is documented in the migration guide, because it is the single most common reason people arrive here.
The temptation foreclosed. The most reliable revenue mechanism in this industry: install, let usage grow for a decade, then price the renewal against the cost of leaving. It works. That is exactly what is wrong with it.
No telemetry
The reasoning. OrdinateDB runs inside OT networks, on plants whose owners have regulatory and commercial reasons to know exactly what crosses their perimeter. The product phones home to nobody, and this is verifiable in source rather than asserted in a privacy policy. Neither does this website.
The temptation foreclosed. "Anonymised usage analytics to inform the roadmap", which is how every phone-home starts, and never how it ends. The only telemetry-free system an OT security review can accept quickly is one with nothing to review.
No proprietary storage format
The reasoning. Your archive should outlive your vendor — including us. The cold tier is documented Apache Parquet in a published layout; DuckDB, Spark and pandas read it with no OrdinateDB process running. Exit is a tested feature, with the test published.
The temptation foreclosed. The quiet v3 "optimised native format" that makes exports need an exporter, and exporters need a licence. Format lock-in is never announced; it is always an optimisation. Ours is foreclosed in writing.
No SSO tax
The reasoning. OIDC single sign-on is in the free core. Charging extra for SSO prices security as a luxury, which is exactly backwards for software that sits on plant networks: the installations least able to pay for the enterprise tier are the ones an attacker would most like to find sharing a local password.
The temptation foreclosed. The most rational upsell in commercial software — SSO is cheap to build and enterprises must have it, which is why the practice is common enough to have earned its own name and a community-maintained wall of shame. Rational, and wrong, and pre-declined.
No invented values
The reasoning. A historian records; it does not imagine. When a collector was down, the gap is a gap — drawn as one, aggregated as one, confessed in pct_good on every result. No interpolation over outages, no filled-in estimates presented at the same rank as measurements, no machine-generated "insights" that cannot cite the points they came from.
The temptation foreclosed. The current decade's version is obvious: a model that fills your gaps and summarises your trends, demoed on data where it happens to be right. An archive whose values are sometimes measurements and sometimes guesses is not an archive with extra features; it is a different, lesser kind of thing.
Why the list is locked
A non-goals list on a marketing page is worth the pixels it is printed on, and no more. Companies change owners; projects change stewards; every commitment above will eventually be inconvenient to somebody with commit access.
So the list does not live on a marketing page. It lives in the project's governance documents, and amending it requires a supermajority of the technical steering committee and a public comment period — the same procedure as a licence change, because it is the same kind of promise. A future steward who wants to meter tags cannot merge that decision; they must propose it, in public, over the objections the process is designed to invite, with this post on the record as the project's considered answer.
That is the actual product here. The archive engine is good, and you should judge it on its evidence. But historians are thirty-year commitments, and what you are really choosing is the set of things your vendor will do to you in year fifteen. Ours is written down, and we have arranged to be unable to change it quietly.