finding(core/contract): the seven semantic sectioning tags are declared layout containers the curated vocabulary has never ruled on, either way

[os-bill]

Filed out-of-scope by the os-dev seat executing objectui#8499 (session session_012W3vMLTFY9SPr2LyxhSeYi). ⛔ Recording only — no domain:*, no grading, no type; that is triage's to produce. ⛔ Not decided on objectui#8499's PR: that card arms AnyComponentSchema for renderers that already run, and curating (or refusing) seven types in the public vocabulary is a separate decision that no ruling on that card covers.

This is the same shape as objectui#8628, one family later, and it is filed for the same reason: the ledger entry that keeps these visible has to name an issue that resolves it.

How they arrived in the population, measured

The derived population objectui#6879 built for its forcing function is

LayoutSchema.options            (the runtime-enumerable zod authoring union,
                                 packages/types/src/zod/layout.zod.ts)
  ∩ ComponentRegistry meta isContainer === true

objectui#8499 adds SemanticElementSchema — one arm whose type is a z.enum over the seven tags packages/components/src/renderers/layout/semantic.tsx registers. Those seven were already registered with isContainer: true (objectui#6764); what changed is that they are now also arms of the layout union, so the intersection above reaches them for the first time.

Measured on this card's branch, the population goes from 7 members to 14:

type	in PUBLIC_BLOCKS?
card · container · flex · grid · stack	yes — the objectui#6764 control set
box	yes — curated by objectui#6879
aspect-ratio	no — objectui#8628
article · aside · footer · header · main · nav · section	no, and no card has ever said why

Each of the seven registers category: 'layout', isContainer: true, and exactly one input (className).

Two further readings, so the decision is not taken on a guess:

• The spec-parity gate cannot see any of them. Object.keys(ComponentPropsMap) carries no entry for any of the seven, so promoting one demands no @objectstack/spec entry and loosens nothing there. (Same answer objectui#6879 measured for box.)
• They are not unvalidated today. renderers/layout/page.tsx builds the JSX-page compiler's manifest from getKnownTypes(), registry-wide with no roster in the path, so validateTree already judges them.

⚠️ Why promoting one is NOT a pure widening — the half that makes this a decision

⛔ CORRECTED 2026-09-09. The version of this section filed with the card gave a mechanism that is false — that curating one of these would delete the tag from every react page. It does not. The measurement and the corrected mechanism are below; a correction comment on this card records the change. ⭐ The decision this card asks for is unchanged, and so is the fact that promotion is not a roster edit — only the consequence to carry is different.

packages/components/src/renderers/layout/semantic.tsx records the consequence at its own registration, and __tests__/container-declaration-census.test.tsx pins the premise:

Safe on the second consumer, and that was MEASURED rather than assumed (objectui#6764): renderers/layout/react-page.tsx drops containers from the kind:'react' JSX scope, but it iterates getPublicConfigs(), and none of these seven tags is in the curated PUBLIC_BLOCKS contract -- so there is no injected Aside/Main/... identifier for this flag to remove. __tests__/container-declaration-census.test.tsx pins that premise, so promoting one of these into the public contract re-opens the question here instead of silently deleting a tag from every react page.

⚠️ That quote is accurate; the inference first drawn from it was not. It describes objectui#6764's direction — declaring containment on a block that is already public removes an injected identifier that existed — and reading it forwards, as a statement about promotion, inverts it.

⇒ Measured, on main:

leg	reading
react-page.tsx:75-77	if (!tag || cfg.isContainer) continue; — it skips every container config, whichever list the config came from
getMeta('main').isContainer / getConfig('main').isContainer	true / true — the seven have carried the flag since objectui#6764
injected identifiers today	46
injected identifiers with main promoted (simulated)	46
Main in either set	absent from both; set difference []
ComponentRegistry references in packages/react-runtime/src	0 (control: 21 React references in the same file)

⇒ A lowercase main in a kind:'react' page is a DOM intrinsic before and after; nothing stops resolving, because there was never an injected identifier and promotion does not create one.

⇒ What promotion actually costs, and it is still enough to make this a decision rather than a roster edit:

1. It widens the AI-authoring vocabulary, sdui.manifest.json and the generated intrinsics — a public-contract surface, permanently.
2. It turns container-declaration-census.test.tsx's "none of the eight is in the curated public contract" pin red by design. That is the pin doing its job — "re-opens the question HERE" is exactly this — and whoever promotes must move that pin deliberately, with the census population's reason restated.

⛔ What it does not cost is any react page.

What the decision looks like

Two coherent answers, and this card needs only one of them chosen:

1. Curate the family (or a named subset) into PUBLIC_BLOCKS Tier B — accepting the vocabulary / manifest / intrinsics widening, and moving container-declaration-census.test.tsx deliberately rather than discovering it red. ⛔ No react-page.tsx change is required; that premise was the erratum.
2. Refuse them on stated merits — e.g. "the curated vocabulary carries one spelling per concept and box plus a className is that spelling; the sectioning tags exist for parsed kind:'html' documents, which reach the renderer without the roster" — and replace the ledger text in apps/console/src/__tests__/public-contract.test.ts with those merits.

Until one is chosen the seven are ledgered in UNCURATED_LAYOUT_CONTAINERS in that file, which is a forcing function rather than an allowlist: they are visible there instead of invisible in a gap, and this issue is what the entries name.

Reproduction

From the repo root, pnpm exec vitest run apps/console/src/__tests__/public-contract.test.ts — the PUBLIC_BLOCKS ↔ the declared layout containers block derives the population and prints it; delete the seven ledger entries and the curates every declared layout container, or records why not case fails by absence.

---

Generated by Claude Code

[os-bill]

⛔ Correction to this card's body — the mechanism it gave for "promotion is not a pure widening" was FALSE.

domain:spec @ objectui seat, session session_012W3vMLTFY9SPr2LyxhSeYi, 2026-09-09. Filed by the same seat that filed the card, on the ceiling re-review's finding at objectui#8499 comment 5597738741. The body has been edited in place; this comment records what changed and why, so the edit is visible in the timeline rather than silent.

What was wrong

The card said: curating one of the seven into PUBLIC_BLOCKS would make react-page.tsx drop that tag from the kind:'react' JSX scope, so a main an author writes in a react page today "would stop resolving".

It does not. Re-measured on main rather than taken on the review's word:

leg	reading
react-page.tsx:75-77	if (!tag || cfg.isContainer) continue; — skips every container config, whichever list it came from
getMeta('main').isContainer / getConfig('main').isContainer	true / true — the seven have carried the flag since objectui#6764
injected identifiers today	46
injected identifiers with main promoted (simulated: push main onto PUBLIC_BLOCKS, re-run the scope loop)	46
Main in either set	absent from both; set difference []
the promoted config's own flag	{"type":"main","isContainer":true} — skipped on the same line an unlisted tag never reaches
ComponentRegistry references in packages/react-runtime/src	0 (control: 21 React references in the same file; its only imports are react and sucrase)

⇒ a lowercase main in a react page is a DOM intrinsic before and after. There was never an injected identifier for the flag to remove, and promotion does not create one.

Where the error came from

The semantic.tsx docblock the card quotes is accurate — and the quote stays in the body. It describes objectui#6764's direction: declaring containment on a block that is already public removes an injected identifier that existed. Reading it forwards, as a statement about promotion, inverts it. That is the whole defect, and it is worth naming precisely because the quote itself is true.

What actually changes on promotion

1. It widens the AI-authoring vocabulary, sdui.manifest.json and the generated intrinsics — a permanent public-contract surface.
2. It turns container-declaration-census.test.tsx's "none of the eight is in the curated public contract" pin red by design — which is precisely what semantic.tsx means by "re-opens the question HERE". Whoever promotes moves that pin deliberately.

⛔ What it does not cost is any react page.

What does NOT change

⭐ The decision this card asks for stands exactly as filed, and so does the framing: promotion is not a roster edit. Both options remain open and neither is weakened — option 1 simply carries a different (and smaller) consequence than the body first claimed, and needs ⛔ no react-page.tsx change. This card stays open; nothing here decides it.

The same correction has been applied to the three other places the false mechanism was published: the shipped ledger string SECTIONING_TAG_UNRULED and the UNCURATED_LAYOUT_CONTAINERS docblock in apps/console/src/__tests__/public-contract.test.ts, and the body of PR #8763 (commit 005409fc1).

---

Generated by Claude Code