os-zhuang · Nov 9, 2026 · Nov 9, 2026 · Nov 9, 2026 · Nov 9, 2026
@@ -1,6 +1,6 @@
 # ADR-0048: Cross-package metadata collision — package-id identity, namespace install gate, package-scoped resolution
 
-**Status**: Revised (2026-06-13) — supersedes the original *per-item collision detection* framing. The runtime guard shipped under the original proposal is **retained as a same-package authoring backstop**; the strategic direction for the app-marketplace era is revised below. · **Superseded in part by [ADR-0146](./0146-studio-designs-setup-administers-one-editor-per-item-paired-customize-package.md)** (2026-10-31) — §3.3's sentence that a context-free read "is best-effort: it returns the first match": a read of a name a code package ships has an owner — that package, or the declared override its paired `<base>.custom` publishes — and a same-named item of a package outside that closure is served only under its own address (ADR-0146 D8); and §3.4's coexistence "by construction", for a base's dependency closure only: the paired `<base>.custom`'s item bearing a base item's name is a declared override or is refused at publish — declare or rename — and any other dependent package's item bearing a base item's name is refused at publish — rename; only the pair declares (ADR-0146 D2, ruled 2026-10-31 on #28364 and #28390). Packages with no dependency relation keep §3.3 and §3.4 as written; qualify-to-cross, the same-package overwrite and the 2026-10-08 addendum stand. · **Amended** (2026-11-07, [#30794](https://github.com/objectstack-ai/objectstack/issues/30794)): §3.5 — rename-on-install ships for **unmanaged** copies, through the one rename engine `duplicatePackage` applies; a managed install is never renamed and §3.2 stands for it as written. See the dated note under §3.5. · **Amended by [ADR-0150](./0150-behavioural-metadata-addressed-by-package-objects-keep-the-literal-prefix.md)** (proposed 2026-11-08): the 2026-11-06 addendum's one-holder refusal retires type by type as each type's readers carry the package; it stays for platform-requested bare names and, as ruled (ADR-0150 Q2 = A, record [`6724050556`](https://github.com/objectstack-ai/objectstack/issues/31527#issuecomment-6724050556) on #31527), for positions. This line takes effect when ADR-0150 is accepted.
+**Status**: Revised (2026-06-13) — supersedes the original *per-item collision detection* framing. The runtime guard shipped under the original proposal is **retained as a same-package authoring backstop**; the strategic direction for the app-marketplace era is revised below. · **Superseded in part by [ADR-0146](./0146-studio-designs-setup-administers-one-editor-per-item-paired-customize-package.md)** (2026-10-31) — §3.3's sentence that a context-free read "is best-effort: it returns the first match": a read of a name a code package ships has an owner — that package, or the declared override its paired `<base>.custom` publishes — and a same-named item of a package outside that closure is served only under its own address (ADR-0146 D8); and §3.4's coexistence "by construction", for a base's dependency closure only: the paired `<base>.custom`'s item bearing a base item's name is a declared override or is refused at publish — declare or rename — and any other dependent package's item bearing a base item's name is refused at publish — rename; only the pair declares (ADR-0146 D2, ruled 2026-10-31 on #28364 and #28390). Packages with no dependency relation keep §3.3 and §3.4 as written; qualify-to-cross, the same-package overwrite and the 2026-10-08 addendum stand. · **Amended** (2026-11-07, [#30794](https://github.com/objectstack-ai/objectstack/issues/30794)): §3.5 — rename-on-install ships for **unmanaged** copies, through the one rename engine `duplicatePackage` applies; a managed install is never renamed and §3.2 stands for it as written. See the dated note under §3.5. · **Amended by [ADR-0150](./0150-behavioural-metadata-addressed-by-package-objects-keep-the-literal-prefix.md)** (proposed 2026-11-08): the 2026-11-06 addendum's one-holder refusal retires type by type as each type's readers carry the package; it stays for platform-requested bare names and, as ruled (ADR-0150 Q2 = A, record [`6724050556`](https://github.com/objectstack-ai/objectstack/issues/31527#issuecomment-6724050556) on #31527), for positions. This line takes effect when ADR-0150 is accepted. · **Amended** (2026-11-09, [#31685](https://github.com/objectstack-ai/objectstack/issues/31685)): §5's authoring-lint bullet, rewritten in place — since ADR-0150 child P ([#31669](https://github.com/objectstack-ai/objectstack/issues/31669)) the bare-name finding of `naming/namespace-prefix` judges only capabilities and positions, and the refusal #30806 phase 3 was to add is withdrawn. §3.4's lint sentence stands as written; see the dated note under §3.4.
 **Deciders**: ObjectStack Protocol Architects
 **Builds on**: [ADR-0003](./0003-package-as-first-class-citizen.md) (package as first-class citizen), [ADR-0005](./0005-metadata-customization-overlay.md) (artifact vs runtime overlay precedence), [ADR-0008](./0008-metadata-repository-and-change-log.md) (metadata repository, `MetaRef` identity), [ADR-0010](./0010-metadata-protection-model.md) (package provenance / `_packageId` stamping)
 **Consumers**: `@objectstack/objectql` (`SchemaRegistry.registerItem`, `ObjectQL.registerApp`, install path), `@objectstack/console` / `objectui` (routing + metadata resolution), package authors, CLI/CI install path, the app marketplace registry.
@@ -290,6 +290,15 @@ package — stays covered by the `naming/namespace-prefix` lint in `os lint`.
 > action bound to an object. See
 > [Addendum (2026-11-06)](#addendum-2026-11-06-bare-names-resolved-without-a-package-hold-one-package--34-narrowed).
 
+> **Note (2026-11-09, [#31685](https://github.com/objectstack-ai/objectstack/issues/31685)) — the
+> lint sentence above still holds.** [ADR-0150](./0150-behavioural-metadata-addressed-by-package-objects-keep-the-literal-prefix.md)
+> child P ([#31669](https://github.com/objectstack-ai/objectstack/issues/31669)) retired the
+> rule's bare-name finding for the types ADR-0150 D1 puts in buckets 2–4, and kept the finding
+> this section relies on (ADR-0150 D9): two items of one type that land on one registry key
+> inside one package — among its apps, pages, dashboards, flows, actions (per `objectName`),
+> reports and datasets (`DUPLICATE_KEYED_TYPES`, `packages/cli/src/commands/lint.ts`) — are
+> still reported. What the rule covers now is in §5.
+
 ### 3.5 Namespace rename-on-install is deferred (non-goal)
 
 Renaming a colliding namespace on install would require rewriting **every**
@@ -381,25 +390,34 @@ alias is optional and out of scope for the phases below.
 Status legend: **[done]** shipped · **[proposed]** not yet built ·
 **[deferred]** out of scope here.
 
-- **[done]** Authoring lint: `naming/namespace-prefix` in `os lint` warns on
-  every bare package-authored top-level name in every manifest collection of a
-  package that declares `manifest.namespace` (a package without one is not
-  judged), and gives the rename as its `fix` (ruling C on #30798, #30804). It
-  does not judge objects (`defineStack` refuses those), views (named for their
-  object), docs (`docs/namespace-prefix`), extensions, translation bundles,
-  fields, `functions`, or an item that declares `overrides`. A name carries the
-  namespace when it is the namespace itself (the namespace-named app per
-  ADR-0019), starts with `<ns>_`, is a dotted `<ns>.<short>` name, the
-  capability shape (maintainer ruling 2026-11-06T11:08Z, recorded in
-  [#30804 comment 6683456244](https://github.com/objectstack-ai/objectstack/issues/30804#issuecomment-6683456244)),
-  or starts with `sys_`. The permission set `guest_portal` is exempt, because
-  the public form doors request it by that bare name (ADR-0138; maintainer
-  ruling 2026-11-06T14:57Z, recorded in
-  [#30804 comment 6687835712](https://github.com/objectstack-ai/objectstack/issues/30804#issuecomment-6687835712);
-  `PLATFORM_REQUESTED_BARE_NAMES`). The rule also reports a name declared
-  twice in one package among its apps, pages, dashboards, flows, actions (per
-  `objectName`), reports and datasets. It is warning-only; the refusal is
-  phase 3 of the same ruling (#30806).
+- **[done]** Authoring lint: `naming/namespace-prefix` in `os lint` reports two
+  warning-only findings over the declarations of one package (amended in place
+  2026-11-09, [#31685](https://github.com/objectstack-ai/objectstack/issues/31685), after
+  ADR-0150 child P, [#31669](https://github.com/objectstack-ai/objectstack/issues/31669)).
+  - **The duplicate finding** (§3.4, kept by ADR-0150 D9): a name declared twice in one
+    package among its apps, pages, dashboards, flows, actions (per `objectName`), reports
+    and datasets (`DUPLICATE_KEYED_TYPES`). It asks for a distinct name, never for a prefix.
+  - **The bare-name finding** judges only the collections whose names keep the literal
+    prefix (`PREFIX_JUDGED_COLLECTIONS`): capabilities, spelled `NS.SHORT` (ADR-0150 D1
+    bucket 1; maintainer ruling 2026-11-06T11:08Z, recorded in
+    [#30804 comment 6683456244](https://github.com/objectstack-ai/objectstack/issues/30804#issuecomment-6683456244)),
+    and positions, one slot per name in the environment (ADR-0150 D7). It judges a package
+    that declares `manifest.namespace`, and gives the rename as its `fix`. A name carries
+    the namespace when it is the namespace itself, starts with `<ns>_`, is a dotted
+    `<ns>.<short>` name, or starts with `sys_`; an item that declares `overrides` is not
+    judged.
+  - **Every other type keeps its short name.** ADR-0150 identifies buckets 2–4 by
+    (package, name) (D1), so child P retired the warning that asked for `<ns>_<short>`
+    there (D9), and a name already prefixed stays valid. The permission set `guest_portal`,
+    exempted by ruling 2026-11-06T14:57Z
+    ([#30804 comment 6687835712](https://github.com/objectstack-ai/objectstack/issues/30804#issuecomment-6687835712);
+    `PLATFORM_REQUESTED_BARE_NAMES`), is no longer judged at all: permission sets are
+    bucket 2.
+  - **No refusal follows.** Phase 3 of ruling C on #30798 (#30806), which was to turn the
+    bare-name finding into a refusal at `defineStack` and the save door, is withdrawn
+    (maintainer ruling A 「转向包寻址」, recorded in
+    [#30806 comment 6721941146](https://github.com/objectstack-ai/objectstack/issues/30806#issuecomment-6721941146);
+    ADR-0150's status line).
 - **[done] Phase 1 — install-time namespace gate.** `NamespaceConflictError` +
   the gate in `SchemaRegistry.installPackage` (refuses a package whose
   `manifest.namespace` is already owned by a different installed package;

@@ -5,7 +5,11 @@ convoy with #12786). This ADR **records an already-enforced, platform-wide decis
 changes no shipped behavior. The contract below has been enforced since the parse path was
 closed by #4001 and is restated by Prime Directive #6 in `AGENTS.md` — what was missing was
 the decision record itself, and that absence is what produced the phantom citation this ADR
-closes (see § History).
+closes (see § History). · **Cross-referenced by
+[ADR-0150](./0150-behavioural-metadata-addressed-by-package-objects-keep-the-literal-prefix.md)**
+(2026-11-09, [#31685](https://github.com/objectstack-ai/objectstack/issues/31685)): objects keep this
+record's literal prefix unchanged, and behavioural metadata is addressed by (package, name) instead
+of carrying it. See the dated note under D4.
 **Deciders**: ObjectStack maintainer (2026-08-29), recording a contract in force since #4001.
 **Relates to**: [ADR-0028](./0028-metadata-naming-and-namespace-isolation.md) (the **Deferred
 target model** pointing the opposite way — reconciled in D4 below),
@@ -82,6 +86,19 @@ design, this ADR is the enforced present. **Until 0028's recorded re-open trigge
 this ADR is the governing record of object naming.** If 0028 is ever adopted, the adopting
 work must supersede this record explicitly (Prime Directive #13) — never drift past it.
 
+> **Note (2026-11-09, [#31685](https://github.com/objectstack-ai/objectstack/issues/31685)) — ADR-0028
+> is re-opened for the behavioural tier only, and this record stays the governing record of object
+> naming.** [ADR-0150](./0150-behavioural-metadata-addressed-by-package-objects-keep-the-literal-prefix.md)
+> (accepted by the maintainer's merge of #31586) supersedes ADR-0028 D1 and D3 in part, for the
+> behavioural and presentation types of its buckets 2–4. Objects are its bucket 1, with fields,
+> object extensions, views, capabilities, docs and an api's path: they keep the literal prefix, for
+> the reason D1 here gives (the name is the table name, read by consumers the platform does not
+> control; ADR-0150 D1). Nothing in D1–D3 changes. What ADR-0150 does not take is ruling C on #30798's
+> extension of D2's literal prefix to every package-authored name: phase 3 of that ruling (#30806),
+> which was to amend this record so, is withdrawn (maintainer ruling A 「转向包寻址」, recorded in
+> [#30806 comment 6721941146](https://github.com/objectstack-ai/objectstack/issues/30806#issuecomment-6721941146)).
+> A flow, an action, a page or a tool keeps a short name, and the platform carries its package.
+
 ---
 
 ## History — why this record exists

@@ -174,6 +174,21 @@ A release artifact MAY carry N package manifests. Everything inside one artifact
 one file and installed in one act. That joint delivery **is** the ownership proof; nothing else
 needs to assert it.
 
+> **Note (2026-11-09, [#31685](https://github.com/objectstack-ai/objectstack/issues/31685)) — what
+> co-ownership does not decide: how a reference between two packages of one artifact resolves.**
+> The CLI's picklist-reference check and the post-seal install judge read the paragraph above (the
+> install judge until E6, #31731) as
+> letting a package resolve a list a sibling package of the same artifact owns, with no declared
+> dependency (`packages/cli/src/utils/picklist-references.ts`). That allowance narrows to declared
+> dependencies: a bare reference resolves from the referring package's own items, then the unique
+> holder in its declared dependency closure (ADR-0141, plus `sys`), else it is refused naming every
+> candidate ([ADR-0150](./0150-behavioural-metadata-addressed-by-package-objects-keep-the-literal-prefix.md)
+> D3). A sibling in the same artifact is in that closure only when the referring package's manifest
+> declares it, and the author's fix is that one `dependencies` entry. Maintainer ruling N1 on
+> #31678, recorded in
+> [#31678 comment 6729820123](https://github.com/objectstack-ai/objectstack/issues/31678#issuecomment-6729820123).
+> D1's co-ownership claim, and the install gate it corrects, are unchanged.
+
 Therefore the install gate's criterion is corrected from *"is this the same package id?"* to
 *"are these co-owners within one artifact?"*. This is not a new permission being granted; it is
 an existing check being asked the question it was always meant to ask. The 2026-08-06 ruling