Filed unassigned and unlabelled for triage to grade and route — I believe this is domain:devx (content/docs/** by the domain table), which is why the drivers seat is not taking it. Split off from #8283 while claiming that card's drivers half.
The ruling
The maintainer ruled on #8283 (2026-08-13, comment 5281698511) — Option 1: the autonumber contract is stated explicitly as unique and monotonic per scope, NOT gapless — any rejected write may consume the number it reserved. Its stated purpose: "stops anyone promising a customer a gapless series."
Its scope line reads "state the contract where an author looks for autonumber behaviour", and then names the SQL driver's an autonumber is an immutable business identifier sentence as the natural neighbour. Those are two different locations.
Measured on origin/main @ 03b5f81
| Location |
Lane |
Status |
packages/drivers/driver-sql/src/sql-driver.ts:4827 (the named neighbour, a TSDoc) |
domain:drivers |
claimed on #8283, in progress |
content/docs/data-modeling/** — fields.mdx, field-types.mdx, field-type-decision-tree.mdx, schema-design.mdx, indexing.mdx, validation-rules.mdx all carry autonumber material |
domain:devx |
this card |
grep -niE 'gapless|no gaps|gap' across content/docs/data-modeling/*.mdx returns only capability-gap language about aggregation operators, plus one unrelated sentence. Nothing on any author-facing page says a rejected write consumes its number.
Why the driver TSDoc is not sufficient on its own
An app author does not read packages/drivers/driver-sql/src/sql-driver.ts. Neither does an AI writing an ObjectStack app — it reads the docs site and the field-type reference. A contract recorded only in driver source is a decision log, not a guardrail: it documents what the platform chose without reaching the person about to promise a finance team a gapless invoice series.
That is the ruling's own fourth prism — "an explicitly documented 'unique, monotonic, not gapless' contract is precisely the shape an AI app-author cannot mis-promise" — and it only holds where the AI app-author actually reads.
Suggested content (not proposing the wording)
The ruling supplies the substance; this card is about placement. It needs the contract sentence (unique + monotonic per scope, gaps possible) plus the plain-language cause the ruling asked for: a rejected write — a unique violation on another field, a validation rule, a throwing beforeInsert — consumes the number it had reserved. ⚠️ Worth stating the corollary explicitly for the finance case, since that is who asks: an autonumber is not a legally gapless document number.
⛔ Not proposing which page is primary — field-types.mdx's auto_number entry and fields.mdx are both plausible anchors, and that is a devx judgement about their docs' structure.
⚠️ Anchor by phrase, not line number — #8283 records that line numbers in this area drift within hours, and this session measured that three times today.
Related: #8283 (the ruling and the drivers half), #8269 (the pathological sibling whose merge left this as the whole remaining question).
Filed unassigned and unlabelled for triage to grade and route — I believe this is
domain:devx(content/docs/**by the domain table), which is why the drivers seat is not taking it. Split off from #8283 while claiming that card's drivers half.The ruling
The maintainer ruled on #8283 (2026-08-13, comment
5281698511) — Option 1: the autonumber contract is stated explicitly as unique and monotonic per scope, NOT gapless — any rejected write may consume the number it reserved. Its stated purpose: "stops anyone promising a customer a gapless series."Its scope line reads "state the contract where an author looks for autonumber behaviour", and then names the SQL driver's
an autonumber is an immutable business identifiersentence as the natural neighbour. Those are two different locations.Measured on
origin/main@03b5f81packages/drivers/driver-sql/src/sql-driver.ts:4827(the named neighbour, a TSDoc)domain:driverscontent/docs/data-modeling/**—fields.mdx,field-types.mdx,field-type-decision-tree.mdx,schema-design.mdx,indexing.mdx,validation-rules.mdxall carry autonumber materialdomain:devxgrep -niE 'gapless|no gaps|gap'acrosscontent/docs/data-modeling/*.mdxreturns only capability-gap language about aggregation operators, plus one unrelated sentence. Nothing on any author-facing page says a rejected write consumes its number.Why the driver TSDoc is not sufficient on its own
An app author does not read
packages/drivers/driver-sql/src/sql-driver.ts. Neither does an AI writing an ObjectStack app — it reads the docs site and the field-type reference. A contract recorded only in driver source is a decision log, not a guardrail: it documents what the platform chose without reaching the person about to promise a finance team a gapless invoice series.That is the ruling's own fourth prism — "an explicitly documented 'unique, monotonic, not gapless' contract is precisely the shape an AI app-author cannot mis-promise" — and it only holds where the AI app-author actually reads.
Suggested content (not proposing the wording)
The ruling supplies the substance; this card is about placement. It needs the contract sentence (unique + monotonic per scope, gaps possible) plus the plain-language cause the ruling asked for: a rejected write — a unique violation on another field, a validation rule, a throwing⚠️ Worth stating the corollary explicitly for the finance case, since that is who asks: an autonumber is not a legally gapless document number.
beforeInsert— consumes the number it had reserved.⛔ Not proposing which page is primary —
field-types.mdx'sauto_numberentry andfields.mdxare both plausible anchors, and that is a devx judgement about their docs' structure.Related: #8283 (the ruling and the drivers half), #8269 (the pathological sibling whose merge left this as the whole remaining question).