docs(spec): quote the registry door drive as it posts, com.example.pkg-a

The install door's residual docblock, its pinning test and the changeset
transcribed the duplicate-id drive in domain-handler-registry.test.ts as
{ id: 'pkg-a', name: 'A', version: '1.0.0' }. That drive stopped posting
that body at PR #19473, which made the door parse the manifest's id leg
and repaired the fixture's id to com.example.pkg-a. The old body is the
REVERSED pin beside it and is answered 400. It had been filed as a 201
residual under clause 1b.

Read at the merge base fdeeea0 and at origin/main 2c1011b (the file
is byte-identical at both): the drive posts
{ id: 'com.example.pkg-a', name: 'A', version: '1.0.0' }, answered 409 and
then 201 on ?overwrite=true. The declaration refuses it on `type` alone.
Clause 1b's "both door drives above" is true of that body, so 1b is
unchanged.

- package-api.zod.ts: the drives paragraph quotes the real body and
  credits both repairs, the version to PR #19326 and the id to PR #19473.
- package-api.test.ts: DOOR_DRIVE_REGISTRY transcribes the real body.
  The control that rested on the stale id now asserts that the registry
  drive parses once `type` is added, like the conflict drive. The old
  `pkg-a` reading is kept, pointed at the old body: it is refused on the
  id alone.
- changeset: the drives sentence names the body the drive posts.

Prose and test only. No schema, accept set, export or runtime change.

Claude-Session: https://claude.ai/code/session_019c3Hi6ZMU1p6m6aA6Bz45d
Co-authored-by: Claude <noreply@anthropic.com>

Author: claude

.changeset/19327-install-door-residual-split.md

@@ -17,8 +17,12 @@ The clause is now split: **1a** (missing `version`) is marked CLOSED by #19326
 and names the door-side pin, and **1b** (missing `type`) stays an open residual.
 The count of five classes is unchanged and still true, because class 1 stays
 open through its `type` half; the count sentence now says so. The paragraph
-that quotes the runtime's two door drives is updated too: the duplicate-id
-drive has posted a `version` since #19326, and the docblock now quotes that body.
+that quotes the runtime's two door drives is updated too. It quoted the
+duplicate-id drive as `{ id: 'pkg-a', name: 'A' }`, but that drive has posted a
+`version` since #19326 and the reverse-domain id `com.example.pkg-a` since
+#19473, and the door answers `400` to `pkg-a`. The docblock now quotes the body
+the drive posts, `{ id: 'com.example.pkg-a', name: 'A', version: '1.0.0' }`,
+which the declaration refuses on `type` alone and the door answers `201`.
 
 ⛔ No behaviour changes. No schema, accept set, export or runtime code moves,
 and the other four residual classes are untouched.

packages/spec/src/api/package-api.test.ts

@@ -921,10 +921,13 @@ describe('#18058 — install contract bound to the live door', () => {
     const DOOR_DRIVE_CONFLICT = { id: 'com.acme.crm', name: 'com.acme.crm', namespace: 'crm', version: '1.0.0' };
     /**
      * The duplicate-id drive in `packages/runtime/src/domain-handler-registry.test.ts` — no `type`.
-     * It posted no `version` either until PR #19326 made the door refuse that half (the docblock's
-     * clause 1a, CLOSED) and gave the drive a `version`; this is the body it posts now.
+     * Two of its keys were repaired, each by the PR that made the door parse that leg: PR #19326
+     * gave it the `version` it lacked (the docblock's clause 1a, CLOSED), and PR #19473 replaced its
+     * id `pkg-a`, which `MANIFEST_ID_PATTERN` refuses. The old body is the REVERSED pin beside the
+     * drive, answered `400`, so ⛔ it is no residual and is not transcribed here. This is the body
+     * the drive posts: `409` first, then `201` on `?overwrite=true`.
      */
-    const DOOR_DRIVE_REGISTRY = { id: 'pkg-a', name: 'A', version: '1.0.0' };
+    const DOOR_DRIVE_REGISTRY = { id: 'com.example.pkg-a', name: 'A', version: '1.0.0' };
 
     it('the namespace-conflict drive is REFUSED — it carries no `type`', () => {
       expect(PackageInstallBodySchema.safeParse(DOOR_DRIVE_CONFLICT).success).toBe(false);
@@ -934,24 +937,22 @@ describe('#18058 — install contract bound to the live door', () => {
       expect(PackageInstallBodySchema.safeParse(DOOR_DRIVE_REGISTRY).success).toBe(false);
     });
 
-    it('the missing keys are what decide it — and since #17534 the registry drive needs its id repaired too', () => {
+    it('the missing `type` is what decides it, for BOTH drives — the registry drive\'s old id is refused on its own', () => {
       // The control that makes the two refusals above a measurement of the
       // MANIFEST's required keys rather than of the bare branch existing at all.
       expect(PackageInstallBodySchema.safeParse({ ...DOOR_DRIVE_CONFLICT, type: 'app' }).success).toBe(true);
-      // ⭐ #17534 moved this half. `ManifestSchema.id` carries
-      // `MANIFEST_ID_PATTERN` now, and `pkg-a` is not reverse-domain notation,
-      // so completing the missing `type` is no longer sufficient for THIS drive —
-      // it stays refused, on the id's shape rather than on an absent key.
-      // ⛔ The remedy is to say that, not to relax the pattern: the drive posts
-      // an id the registry face has always refused to publish.
       const registryKeysCompleted = { ...DOOR_DRIVE_REGISTRY, type: 'app' };
-      expect(PackageInstallBodySchema.safeParse(registryKeysCompleted).success).toBe(false);
-      // Lit control — the id is what decides it now: the same body with a
-      // reverse-domain id parses green, so the refusal above is not the missing
-      // keys coming back.
-      expect(PackageInstallBodySchema.safeParse({
-        ...registryKeysCompleted, id: 'com.acme.pkg-a',
-      }).success).toBe(true);
+      expect(PackageInstallBodySchema.safeParse(registryKeysCompleted).success).toBe(true);
+      // ⭐ Until PR #19473 this half ran the other way. The drive posted `pkg-a`,
+      // which `MANIFEST_ID_PATTERN` has refused since #17534 (PR #18319), so
+      // completing its keys was not sufficient and this case pinned the
+      // refusal. PR #19473 made the door parse the id leg as well, so the door
+      // answers that body `400` (the REVERSED pin beside the drive), and it
+      // repaired the drive's id. ⛔ The old reading is kept, pointed at the old
+      // body: refused on the id alone (the lit control is the green parse just
+      // above, the same body with the drive's current id), so it is no part of
+      // the `201` residual.
+      expect(PackageInstallBodySchema.safeParse({ ...registryKeysCompleted, id: 'pkg-a' }).success).toBe(false);
     });
 
     /** Bodies the door still answers `201` to while this declaration refuses them — the live residual. */
@@ -982,7 +983,8 @@ describe('#18058 — install contract bound to the live door', () => {
       // cited rather than repeated — this package cannot import the door. On
       // that key the declaration and the door agree, so a versionless body in
       // the list above would record a `201` the door no longer answers, which
-      // is what the registry drive's old transcription did.
+      // is what the registry drive's first transcription, `{ id: 'pkg-a',
+      // name: 'A' }`, did.
       // ⛔ Clause 1b stays open: both drives above still carry no `type`.
       for (const residual of DOOR_201_RESIDUALS) {
         const manifest = ('manifest' in residual ? residual.manifest : residual) as { version?: unknown };

packages/spec/src/api/package-api.zod.ts

@@ -478,12 +478,15 @@ export type PackageInstallRequestParsed = z.infer<typeof PackageInstallRequestSc
  *
  * ⚠️ What those two drives post is NOT covered by this branch, and saying so
  * is the point. Measured: `{ id, name: id, namespace, version: '1.0.0' }` and
- * `{ id: 'pkg-a', name: 'A', version: '1.0.0' }` are both refused here
- * (`invalid_union`) because neither carries `type`. The second carried no
- * `version` either until PR #19326, which made the door refuse that half and
- * gave the drive the `version` it lacked (clause 1a below). They are bare in
- * FORM and incomplete in CONTENT — the form is declared, the content is part
- * of the residual below, and they are pinned as REFUSED in
+ * `{ id: 'com.example.pkg-a', name: 'A', version: '1.0.0' }` are both refused
+ * here (`invalid_union`) on `type` alone, and the door answers both `201` (the
+ * second on the `?overwrite=true` limb of its duplicate-id case). The second
+ * was repaired twice, each time by the PR that made the door parse the leg it
+ * broke: PR #19326 gave it the `version` it lacked (clause 1a below), and
+ * PR #19473 replaced its id `pkg-a`, which `MANIFEST_ID_PATTERN` refuses. The
+ * door answers that old body `400` now, so it is no part of the residual. They
+ * are bare in FORM and incomplete in CONTENT — the form is declared, the
+ * content is part of the residual below, and they are pinned as REFUSED in
  * `package-api.test.ts` rather than dressed up as green fixtures.
  *
  * ## The two branches are disjoint — but only ONE of them is closed