fix(service-messaging): stamp organization_id on sys_http_delivery rows so the redeliver() cross-organization wall excludes other tenants' rows

[os-steve]

Fixes#13546

What breaks today

sys_http_delivery is tenant-scoped and redeliver() — its one request-reachable door (POST /api/v1/webhooks/redeliver, open to any authenticated user) — deliberately scopes by the caller's organization (#10740). But the enqueue door never stamped organization_id, and the SQL driver's tenant term is (organization_id = :tenantId OR organization_id IS NULL) — a deliberate global-row fail-open. So 100% of rows landed in the NULL arm: visible to, and replayable by, every organization on a walled deployment. Declared, not enforced; this PR restores the declared contract. The driver's tenant term is untouched — removing the IS NULL arm would hide true global rows from every tenant, a different defect.

The repair — mirror of the sibling that already received it

The notification outbox got exactly this fix (EnqueueDeliveryInput.organizationId, SqlOutbox.enqueue writing organization_id: input.organizationId ?? null). This PR gives the HTTP outbox the same shape, end to end:

• Contract (http-outbox.ts): EnqueueHttpInput gains an organizationId?: string member — inherited by UndeliverableHttpInput, so parked rows are tenant-stamped too — and HttpDelivery surfaces it on read-back.
• SQL impl (sql-http-outbox.ts): the shared insert() (reached from both enqueue() and recordUndeliverable()) writes organization_id: input.organizationId ?? null, the same line the sibling writes; toDelivery() maps it back (NULL → absent).
• Memory impl (memory-http-outbox.ts): stores the same field — and, now that its rows carry a tenant, its redeliver() honors the obligation its own tenant-audit, the update half: ack is a dispatcher sweep but redeliver is request-reachable — two sites on one object with OPPOSITE classifications #10740 note recorded ("a future memory implementation that DOES store a tenant owes the predicate here"), with the driver's exact semantics: another organization's row is invisible (RESOURCE_NOT_FOUND, never an existence oracle), an org-less row stays a global row any tenant may replay, a tenant-less caller stays unscoped.
• Producer 1 — flow http node (http-nodes.ts, durable mode): threads the run's acting organization, AutomationContext.tenantId — the same source and the same no-fallback rule as the notify node's [finding] sys_inbox_message/sys_notification/sys_email 等平台表从不写 organization_id(存量与新增行 100% null)——请确认多组织语义是否设计如此 #11303 repair in the same directory. When a run has none, it warns loudly (fail-loud, not fail-closed, not fail-guess) and enqueues org-less.
• Producer 2 — webhook auto-enqueuer (auto-enqueuer.ts): stamps each delivery with its subscription's own organization (sys_webhook is organization-scoped, Five more instances of the #8323 class: admin- and user-authored names on tenant-scoped objects still carry installation-wide unique indexes #8554, so the row carries the tenant that authored it). There is no request context on this fire-and-forget path; the subscription row is the one honest source. Org-less subscriptions enqueue org-less, silently — a per-event warn here would fire on every record write of a single-posture install.

Forward-stamping only. Existing NULL rows are untouched; their disposition is the separate decision the card names. No backfill, no driver changes, no schema changes (organization_id is kernel-provisioned on every object).

Optional vs required — argued, not assumed

The new member is optional (organizationId?: string). Weighed explicitly against the required-but-undefined-able shape this same file uses for RedeliverOptions.tenantId:

• For optional: it mirrors the sibling notification outbox's EnqueueDeliveryInput.organizationId and EmitInput.organizationId — one convention across the two outboxes of one package (a second convention here would be this card's defect repeated). It is non-breaking for the published surface. And org-less enqueues are legitimate: a single-posture install and every stack before its first organization must keep enqueuing.
• For required (organizationId: string | undefined): nothing tells a future producer it forgot one — the exact gap that produced this bug. That is a real cost. The mitigations shipped here: the producer population of this seam is enumerated (two, both repaired in this PR), the flow node warns loudly on an org-less multi-org run, and the driver's auditMissingTenant line still reports unscoped writes on multi-tenant boots. RedeliverOptions.tenantId went required because its callers are arbitrary request routes; this seam's callers are the two service-internal producers.

If the contract reviewer prefers the required shape, the change is mechanical (both producers already always spell the key or spread it conditionally) but it becomes a compile-breaking widening of a published input.

Write-surface enumeration (the triage's must-answer)

Is SqlHttpOutbox.insert the only door? Enumerated, not exemplified, at origin/main (ff37576):

• Production constructors of an IHttpOutbox: exactly one — messaging-service-plugin.ts:310 (new SqlHttpOutbox(engine, ...), default object name). MemoryHttpOutbox is constructed only by tests.
• Row-creating engine calls on the object: exactly one INSERT site, SqlHttpOutbox.insert, reached from enqueue() and recordUndeliverable() — both covered by this stamp. No bulkCreate/upsert/seed writes anywhere (grep -rn "insert(SYS_HTTP_DELIVERY|'sys_http_delivery'" over packages/ apps/: only the four test files that build their own rows).
• The remaining writes (claim/ack/redeliver UPDATEs) create no rows and never touch the tenant column; userActions declares create/edit/delete false and the object is managedBy: 'engine-owned'.

Verification

• New pins: packages/services/service-messaging/src/http-outbox-organization.test.ts (9 — SQL stamp verbatim / explicit-NULL normalization / parked-row stamp / read-back mapping; memory parity + all three arms of the redeliver predicate, ADR-0112 code asserted on the refusal); http-nodes.test.ts (+2 — tenant threaded verbatim; over-denial control: org-less run still enqueues, key absent, warns); auto-enqueuer.test.ts (+3 — per-record, org-less control, bulk path).
• Targeted suites at head sha e2ecb71: service-messaging 7 files / 48 tests pass (incl. the tenant-audit and parked-record integration suites), service-automation 4 files / 34 pass, plugin-webhooks 3 files / 40 pass. typecheck green for service-messaging + plugin-webhooks (service-automation declares none); --listFiles confirms the new test file is inside the messaging typecheck.
• Reverse verification (committed state, restore proven by blob hash vs HEAD): plugin-webhooks resolves @objectstack/service-messaging via dist/index.d.ts (tsc --listFiles); mutating the producer key to organizationIdX reddens typecheck with TS2561 naming organizationId as the fix — proving the rebuilt .d.ts carries the member and the literal is enforced against it. Restore verified byte-identical (worktree blob = HEAD blob 7f52692).
• Gates at e2ecb71: all 14 path-derived families green except check:dual-build-cjs-loads, which is NOT MEASURED locally (PREREQUISITE NOT MET — it reads every package's dist and 41+ untouched packages are unbuilt here; CI's Build Core owns the full run). Declared narrowing: the gate's own population is every published package's built entry (quoted from its error output); this diff touches 3 packages and no build/exports config of any other, so untouched packages' verdicts cannot move; the 3 touched packages were rebuilt and their CJS entries require() clean (3/3 probed).

Generated by Claude Code

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 3 package(s): @objectstack/plugin-webhooks, @objectstack/service-automation, @objectstack/service-messaging, touching 13 documentable anchor(s).

28 hand-written doc(s) name something this change touched — list omitted above 15 rows. Re-derive on the tree named below: node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0.

⛔ 5 release-owned page(s) also affected — read-only, see AGENTS.md Documentation Guardrails.

What this run could not see

• 1 cross-cutting symbol(s) contributed no route anchor: organizationId (4 routes)
• 2 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 47 of 219 client-bound route-ledger rows — the other 172 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 172: 14 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 56 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 102 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.

Coarse fallback — 6 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 214ce1a01bdf23e06de3ef9a877c3a36013531fd — the merge of head 479e6315071c3b8cef710d46da7e64b4d3a160c1 into base 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 214ce1a01bdf23e06de3ef9a877c3a36013531fd && git checkout 214ce1a01bdf23e06de3ef9a877c3a36013531fd
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 479e6315071c3b8cef710d46da7e64b4d3a160c1 && git checkout -B drift-repro 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 && git merge --no-ff 479e6315071c3b8cef710d46da7e64b4d3a160c1
node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[os-steve]

Fixes#13546

What breaks today

sys_http_delivery is tenant-scoped and redeliver() — its one request-reachable door (POST /api/v1/webhooks/redeliver, open to any authenticated user) — deliberately scopes by the caller's organization (#10740). But the enqueue door never stamped organization_id, and the SQL driver's tenant term is (organization_id = :tenantId OR organization_id IS NULL) — a deliberate global-row fail-open. So 100% of rows landed in the NULL arm: visible to, and replayable by, every organization on a walled deployment. Declared, not enforced; this PR restores the declared contract. The driver's tenant term is untouched — removing the IS NULL arm would hide true global rows from every tenant, a different defect.

The repair — mirror of the sibling that already received it

The notification outbox got exactly this fix (EnqueueDeliveryInput.organizationId, SqlOutbox.enqueue writing organization_id: input.organizationId ?? null). This PR gives the HTTP outbox the same shape, end to end:

• Contract (http-outbox.ts): EnqueueHttpInput gains an organizationId?: string member — inherited by UndeliverableHttpInput, so parked rows are tenant-stamped too — and HttpDelivery surfaces it on read-back.
• SQL impl (sql-http-outbox.ts): the shared insert() (reached from both enqueue() and recordUndeliverable()) writes organization_id: input.organizationId ?? null, the same line the sibling writes; toDelivery() maps it back (NULL → absent).
• Memory impl (memory-http-outbox.ts): stores the same field — and, now that its rows carry a tenant, its redeliver() honors the obligation its own tenant-audit, the update half: ack is a dispatcher sweep but redeliver is request-reachable — two sites on one object with OPPOSITE classifications #10740 note recorded ("a future memory implementation that DOES store a tenant owes the predicate here"), with the driver's exact semantics: another organization's row is invisible (RESOURCE_NOT_FOUND, never an existence oracle), an org-less row stays a global row any tenant may replay, a tenant-less caller stays unscoped.
• Producer 1 — flow http node (http-nodes.ts, durable mode): threads the run's acting organization, AutomationContext.tenantId — the same source and the same no-fallback rule as the notify node's [finding] sys_inbox_message/sys_notification/sys_email 等平台表从不写 organization_id(存量与新增行 100% null)——请确认多组织语义是否设计如此 #11303 repair in the same directory. When a run has none, it warns loudly (fail-loud, not fail-closed, not fail-guess) and enqueues org-less.
• Producer 2 — webhook auto-enqueuer (auto-enqueuer.ts): stamps each delivery with its subscription's own organization (sys_webhook is organization-scoped, Five more instances of the #8323 class: admin- and user-authored names on tenant-scoped objects still carry installation-wide unique indexes #8554, so the row carries the tenant that authored it). There is no request context on this fire-and-forget path; the subscription row is the one honest source. Org-less subscriptions enqueue org-less, silently — a per-event warn here would fire on every record write of a single-posture install.

Forward-stamping only. Existing NULL rows are untouched; their disposition is the separate decision the card names. No backfill, no driver changes, no schema changes (organization_id is kernel-provisioned on every object).

Optional vs required — argued, not assumed

The new member is optional (organizationId?: string). Weighed explicitly against the required-but-undefined-able shape this same file uses for RedeliverOptions.tenantId:

• For optional: it mirrors the sibling notification outbox's EnqueueDeliveryInput.organizationId and EmitInput.organizationId — one convention across the two outboxes of one package (a second convention here would be this card's defect repeated). It is non-breaking for the published surface. And org-less enqueues are legitimate: a single-posture install and every stack before its first organization must keep enqueuing.
• For required (organizationId: string | undefined): nothing tells a future producer it forgot one — the exact gap that produced this bug. That is a real cost. The mitigations shipped here: the producer population of this seam is enumerated (two, both repaired in this PR), the flow node warns loudly on an org-less multi-org run, and the driver's auditMissingTenant line still reports unscoped writes on multi-tenant boots. RedeliverOptions.tenantId went required because its callers are arbitrary request routes; this seam's callers are the two service-internal producers.

If the contract reviewer prefers the required shape, the change is mechanical (both producers already always spell the key or spread it conditionally) but it becomes a compile-breaking widening of a published input.

Write-surface enumeration (the triage's must-answer)

Is SqlHttpOutbox.insert the only door? Enumerated, not exemplified, at origin/main (ff37576):

• Production constructors of an IHttpOutbox: exactly one — messaging-service-plugin.ts:310 (new SqlHttpOutbox(engine, ...), default object name). MemoryHttpOutbox is constructed only by tests.
• Row-creating engine calls on the object: exactly one INSERT site, SqlHttpOutbox.insert, reached from enqueue() and recordUndeliverable() — both covered by this stamp. No bulkCreate/upsert/seed writes anywhere (grep -rn "insert(SYS_HTTP_DELIVERY|'sys_http_delivery'" over packages/ apps/: only the four test files that build their own rows).
• The remaining writes (claim/ack/redeliver UPDATEs) create no rows and never touch the tenant column; userActions declares create/edit/delete false and the object is managedBy: 'engine-owned'.

Verification

• New pins: packages/services/service-messaging/src/http-outbox-organization.test.ts (9 — SQL stamp verbatim / explicit-NULL normalization / parked-row stamp / read-back mapping; memory parity + all three arms of the redeliver predicate, ADR-0112 code asserted on the refusal); http-nodes.test.ts (+2 — tenant threaded verbatim; over-denial control: org-less run still enqueues, key absent, warns); auto-enqueuer.test.ts (+3 — per-record, org-less control, bulk path).
• Targeted suites at head sha e2ecb71: service-messaging 7 files / 48 tests pass (incl. the tenant-audit and parked-record integration suites), service-automation 4 files / 34 pass, plugin-webhooks 3 files / 40 pass. typecheck green for service-messaging + plugin-webhooks (service-automation declares none); --listFiles confirms the new test file is inside the messaging typecheck.
• Reverse verification (committed state, restore proven by blob hash vs HEAD): plugin-webhooks resolves @objectstack/service-messaging via dist/index.d.ts (tsc --listFiles); mutating the producer key to organizationIdX reddens typecheck with TS2561 naming organizationId as the fix — proving the rebuilt .d.ts carries the member and the literal is enforced against it. Restore verified byte-identical (worktree blob = HEAD blob 7f52692).
• Gates at e2ecb71: all 14 path-derived families green except check:dual-build-cjs-loads, which is NOT MEASURED locally (PREREQUISITE NOT MET — it reads every package's dist and 41+ untouched packages are unbuilt here; CI's Build Core owns the full run). Declared narrowing: the gate's own population is every published package's built entry (quoted from its error output); this diff touches 3 packages and no build/exports config of any other, so untouched packages' verdicts cannot move; the 3 touched packages were rebuilt and their CJS entries require() clean (3/3 probed).

Generated by Claude Code

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 3 package(s): @objectstack/plugin-webhooks, @objectstack/service-automation, @objectstack/service-messaging, touching 13 documentable anchor(s).

28 hand-written doc(s) name something this change touched — list omitted above 15 rows. Re-derive on the tree named below: node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0.

⛔ 5 release-owned page(s) also affected — read-only, see AGENTS.md Documentation Guardrails.

What this run could not see

• 1 cross-cutting symbol(s) contributed no route anchor: organizationId (4 routes)
• 2 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 47 of 219 client-bound route-ledger rows — the other 172 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 172: 14 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 56 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 102 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.

Coarse fallback — 6 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 214ce1a01bdf23e06de3ef9a877c3a36013531fd — the merge of head 479e6315071c3b8cef710d46da7e64b4d3a160c1 into base 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 214ce1a01bdf23e06de3ef9a877c3a36013531fd && git checkout 214ce1a01bdf23e06de3ef9a877c3a36013531fd
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 479e6315071c3b8cef710d46da7e64b4d3a160c1 && git checkout -B drift-repro 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 && git merge --no-ff 479e6315071c3b8cef710d46da7e64b4d3a160c1
node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[os-steve]

Fixes#13546

What breaks today

sys_http_delivery is tenant-scoped and redeliver() — its one request-reachable door (POST /api/v1/webhooks/redeliver, open to any authenticated user) — deliberately scopes by the caller's organization (#10740). But the enqueue door never stamped organization_id, and the SQL driver's tenant term is (organization_id = :tenantId OR organization_id IS NULL) — a deliberate global-row fail-open. So 100% of rows landed in the NULL arm: visible to, and replayable by, every organization on a walled deployment. Declared, not enforced; this PR restores the declared contract. The driver's tenant term is untouched — removing the IS NULL arm would hide true global rows from every tenant, a different defect.

The repair — mirror of the sibling that already received it

The notification outbox got exactly this fix (EnqueueDeliveryInput.organizationId, SqlOutbox.enqueue writing organization_id: input.organizationId ?? null). This PR gives the HTTP outbox the same shape, end to end:

• Contract (http-outbox.ts): EnqueueHttpInput gains an organizationId?: string member — inherited by UndeliverableHttpInput, so parked rows are tenant-stamped too — and HttpDelivery surfaces it on read-back.
• SQL impl (sql-http-outbox.ts): the shared insert() (reached from both enqueue() and recordUndeliverable()) writes organization_id: input.organizationId ?? null, the same line the sibling writes; toDelivery() maps it back (NULL → absent).
• Memory impl (memory-http-outbox.ts): stores the same field — and, now that its rows carry a tenant, its redeliver() honors the obligation its own tenant-audit, the update half: ack is a dispatcher sweep but redeliver is request-reachable — two sites on one object with OPPOSITE classifications #10740 note recorded ("a future memory implementation that DOES store a tenant owes the predicate here"), with the driver's exact semantics: another organization's row is invisible (RESOURCE_NOT_FOUND, never an existence oracle), an org-less row stays a global row any tenant may replay, a tenant-less caller stays unscoped.
• Producer 1 — flow http node (http-nodes.ts, durable mode): threads the run's acting organization, AutomationContext.tenantId — the same source and the same no-fallback rule as the notify node's [finding] sys_inbox_message/sys_notification/sys_email 等平台表从不写 organization_id(存量与新增行 100% null)——请确认多组织语义是否设计如此 #11303 repair in the same directory. When a run has none, it warns loudly (fail-loud, not fail-closed, not fail-guess) and enqueues org-less.
• Producer 2 — webhook auto-enqueuer (auto-enqueuer.ts): stamps each delivery with its subscription's own organization (sys_webhook is organization-scoped, Five more instances of the #8323 class: admin- and user-authored names on tenant-scoped objects still carry installation-wide unique indexes #8554, so the row carries the tenant that authored it). There is no request context on this fire-and-forget path; the subscription row is the one honest source. Org-less subscriptions enqueue org-less, silently — a per-event warn here would fire on every record write of a single-posture install.

Forward-stamping only. Existing NULL rows are untouched; their disposition is the separate decision the card names. No backfill, no driver changes, no schema changes (organization_id is kernel-provisioned on every object).

Optional vs required — argued, not assumed

The new member is optional (organizationId?: string). Weighed explicitly against the required-but-undefined-able shape this same file uses for RedeliverOptions.tenantId:

• For optional: it mirrors the sibling notification outbox's EnqueueDeliveryInput.organizationId and EmitInput.organizationId — one convention across the two outboxes of one package (a second convention here would be this card's defect repeated). It is non-breaking for the published surface. And org-less enqueues are legitimate: a single-posture install and every stack before its first organization must keep enqueuing.
• For required (organizationId: string | undefined): nothing tells a future producer it forgot one — the exact gap that produced this bug. That is a real cost. The mitigations shipped here: the producer population of this seam is enumerated (two, both repaired in this PR), the flow node warns loudly on an org-less multi-org run, and the driver's auditMissingTenant line still reports unscoped writes on multi-tenant boots. RedeliverOptions.tenantId went required because its callers are arbitrary request routes; this seam's callers are the two service-internal producers.

If the contract reviewer prefers the required shape, the change is mechanical (both producers already always spell the key or spread it conditionally) but it becomes a compile-breaking widening of a published input.

Write-surface enumeration (the triage's must-answer)

Is SqlHttpOutbox.insert the only door? Enumerated, not exemplified, at origin/main (ff37576):

• Production constructors of an IHttpOutbox: exactly one — messaging-service-plugin.ts:310 (new SqlHttpOutbox(engine, ...), default object name). MemoryHttpOutbox is constructed only by tests.
• Row-creating engine calls on the object: exactly one INSERT site, SqlHttpOutbox.insert, reached from enqueue() and recordUndeliverable() — both covered by this stamp. No bulkCreate/upsert/seed writes anywhere (grep -rn "insert(SYS_HTTP_DELIVERY|'sys_http_delivery'" over packages/ apps/: only the four test files that build their own rows).
• The remaining writes (claim/ack/redeliver UPDATEs) create no rows and never touch the tenant column; userActions declares create/edit/delete false and the object is managedBy: 'engine-owned'.

Verification

• New pins: packages/services/service-messaging/src/http-outbox-organization.test.ts (9 — SQL stamp verbatim / explicit-NULL normalization / parked-row stamp / read-back mapping; memory parity + all three arms of the redeliver predicate, ADR-0112 code asserted on the refusal); http-nodes.test.ts (+2 — tenant threaded verbatim; over-denial control: org-less run still enqueues, key absent, warns); auto-enqueuer.test.ts (+3 — per-record, org-less control, bulk path).
• Targeted suites at head sha e2ecb71: service-messaging 7 files / 48 tests pass (incl. the tenant-audit and parked-record integration suites), service-automation 4 files / 34 pass, plugin-webhooks 3 files / 40 pass. typecheck green for service-messaging + plugin-webhooks (service-automation declares none); --listFiles confirms the new test file is inside the messaging typecheck.
• Reverse verification (committed state, restore proven by blob hash vs HEAD): plugin-webhooks resolves @objectstack/service-messaging via dist/index.d.ts (tsc --listFiles); mutating the producer key to organizationIdX reddens typecheck with TS2561 naming organizationId as the fix — proving the rebuilt .d.ts carries the member and the literal is enforced against it. Restore verified byte-identical (worktree blob = HEAD blob 7f52692).
• Gates at e2ecb71: all 14 path-derived families green except check:dual-build-cjs-loads, which is NOT MEASURED locally (PREREQUISITE NOT MET — it reads every package's dist and 41+ untouched packages are unbuilt here; CI's Build Core owns the full run). Declared narrowing: the gate's own population is every published package's built entry (quoted from its error output); this diff touches 3 packages and no build/exports config of any other, so untouched packages' verdicts cannot move; the 3 touched packages were rebuilt and their CJS entries require() clean (3/3 probed).

Generated by Claude Code

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 3 package(s): @objectstack/plugin-webhooks, @objectstack/service-automation, @objectstack/service-messaging, touching 13 documentable anchor(s).

28 hand-written doc(s) name something this change touched — list omitted above 15 rows. Re-derive on the tree named below: node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0.

⛔ 5 release-owned page(s) also affected — read-only, see AGENTS.md Documentation Guardrails.

What this run could not see

• 1 cross-cutting symbol(s) contributed no route anchor: organizationId (4 routes)
• 2 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 47 of 219 client-bound route-ledger rows — the other 172 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 172: 14 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 56 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 102 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.

Coarse fallback — 6 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 214ce1a01bdf23e06de3ef9a877c3a36013531fd — the merge of head 479e6315071c3b8cef710d46da7e64b4d3a160c1 into base 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 214ce1a01bdf23e06de3ef9a877c3a36013531fd && git checkout 214ce1a01bdf23e06de3ef9a877c3a36013531fd
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 479e6315071c3b8cef710d46da7e64b4d3a160c1 && git checkout -B drift-repro 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 && git merge --no-ff 479e6315071c3b8cef710d46da7e64b4d3a160c1
node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[os-steve]

Fixes#13546

What breaks today

sys_http_delivery is tenant-scoped and redeliver() — its one request-reachable door (POST /api/v1/webhooks/redeliver, open to any authenticated user) — deliberately scopes by the caller's organization (#10740). But the enqueue door never stamped organization_id, and the SQL driver's tenant term is (organization_id = :tenantId OR organization_id IS NULL) — a deliberate global-row fail-open. So 100% of rows landed in the NULL arm: visible to, and replayable by, every organization on a walled deployment. Declared, not enforced; this PR restores the declared contract. The driver's tenant term is untouched — removing the IS NULL arm would hide true global rows from every tenant, a different defect.

The repair — mirror of the sibling that already received it

The notification outbox got exactly this fix (EnqueueDeliveryInput.organizationId, SqlOutbox.enqueue writing organization_id: input.organizationId ?? null). This PR gives the HTTP outbox the same shape, end to end:

• Contract (http-outbox.ts): EnqueueHttpInput gains an organizationId?: string member — inherited by UndeliverableHttpInput, so parked rows are tenant-stamped too — and HttpDelivery surfaces it on read-back.
• SQL impl (sql-http-outbox.ts): the shared insert() (reached from both enqueue() and recordUndeliverable()) writes organization_id: input.organizationId ?? null, the same line the sibling writes; toDelivery() maps it back (NULL → absent).
• Memory impl (memory-http-outbox.ts): stores the same field — and, now that its rows carry a tenant, its redeliver() honors the obligation its own tenant-audit, the update half: ack is a dispatcher sweep but redeliver is request-reachable — two sites on one object with OPPOSITE classifications #10740 note recorded ("a future memory implementation that DOES store a tenant owes the predicate here"), with the driver's exact semantics: another organization's row is invisible (RESOURCE_NOT_FOUND, never an existence oracle), an org-less row stays a global row any tenant may replay, a tenant-less caller stays unscoped.
• Producer 1 — flow http node (http-nodes.ts, durable mode): threads the run's acting organization, AutomationContext.tenantId — the same source and the same no-fallback rule as the notify node's [finding] sys_inbox_message/sys_notification/sys_email 等平台表从不写 organization_id(存量与新增行 100% null)——请确认多组织语义是否设计如此 #11303 repair in the same directory. When a run has none, it warns loudly (fail-loud, not fail-closed, not fail-guess) and enqueues org-less.
• Producer 2 — webhook auto-enqueuer (auto-enqueuer.ts): stamps each delivery with its subscription's own organization (sys_webhook is organization-scoped, Five more instances of the #8323 class: admin- and user-authored names on tenant-scoped objects still carry installation-wide unique indexes #8554, so the row carries the tenant that authored it). There is no request context on this fire-and-forget path; the subscription row is the one honest source. Org-less subscriptions enqueue org-less, silently — a per-event warn here would fire on every record write of a single-posture install.

Forward-stamping only. Existing NULL rows are untouched; their disposition is the separate decision the card names. No backfill, no driver changes, no schema changes (organization_id is kernel-provisioned on every object).

Optional vs required — argued, not assumed

The new member is optional (organizationId?: string). Weighed explicitly against the required-but-undefined-able shape this same file uses for RedeliverOptions.tenantId:

• For optional: it mirrors the sibling notification outbox's EnqueueDeliveryInput.organizationId and EmitInput.organizationId — one convention across the two outboxes of one package (a second convention here would be this card's defect repeated). It is non-breaking for the published surface. And org-less enqueues are legitimate: a single-posture install and every stack before its first organization must keep enqueuing.
• For required (organizationId: string | undefined): nothing tells a future producer it forgot one — the exact gap that produced this bug. That is a real cost. The mitigations shipped here: the producer population of this seam is enumerated (two, both repaired in this PR), the flow node warns loudly on an org-less multi-org run, and the driver's auditMissingTenant line still reports unscoped writes on multi-tenant boots. RedeliverOptions.tenantId went required because its callers are arbitrary request routes; this seam's callers are the two service-internal producers.

If the contract reviewer prefers the required shape, the change is mechanical (both producers already always spell the key or spread it conditionally) but it becomes a compile-breaking widening of a published input.

Write-surface enumeration (the triage's must-answer)

Is SqlHttpOutbox.insert the only door? Enumerated, not exemplified, at origin/main (ff37576):

• Production constructors of an IHttpOutbox: exactly one — messaging-service-plugin.ts:310 (new SqlHttpOutbox(engine, ...), default object name). MemoryHttpOutbox is constructed only by tests.
• Row-creating engine calls on the object: exactly one INSERT site, SqlHttpOutbox.insert, reached from enqueue() and recordUndeliverable() — both covered by this stamp. No bulkCreate/upsert/seed writes anywhere (grep -rn "insert(SYS_HTTP_DELIVERY|'sys_http_delivery'" over packages/ apps/: only the four test files that build their own rows).
• The remaining writes (claim/ack/redeliver UPDATEs) create no rows and never touch the tenant column; userActions declares create/edit/delete false and the object is managedBy: 'engine-owned'.

Verification

• New pins: packages/services/service-messaging/src/http-outbox-organization.test.ts (9 — SQL stamp verbatim / explicit-NULL normalization / parked-row stamp / read-back mapping; memory parity + all three arms of the redeliver predicate, ADR-0112 code asserted on the refusal); http-nodes.test.ts (+2 — tenant threaded verbatim; over-denial control: org-less run still enqueues, key absent, warns); auto-enqueuer.test.ts (+3 — per-record, org-less control, bulk path).
• Targeted suites at head sha e2ecb71: service-messaging 7 files / 48 tests pass (incl. the tenant-audit and parked-record integration suites), service-automation 4 files / 34 pass, plugin-webhooks 3 files / 40 pass. typecheck green for service-messaging + plugin-webhooks (service-automation declares none); --listFiles confirms the new test file is inside the messaging typecheck.
• Reverse verification (committed state, restore proven by blob hash vs HEAD): plugin-webhooks resolves @objectstack/service-messaging via dist/index.d.ts (tsc --listFiles); mutating the producer key to organizationIdX reddens typecheck with TS2561 naming organizationId as the fix — proving the rebuilt .d.ts carries the member and the literal is enforced against it. Restore verified byte-identical (worktree blob = HEAD blob 7f52692).
• Gates at e2ecb71: all 14 path-derived families green except check:dual-build-cjs-loads, which is NOT MEASURED locally (PREREQUISITE NOT MET — it reads every package's dist and 41+ untouched packages are unbuilt here; CI's Build Core owns the full run). Declared narrowing: the gate's own population is every published package's built entry (quoted from its error output); this diff touches 3 packages and no build/exports config of any other, so untouched packages' verdicts cannot move; the 3 touched packages were rebuilt and their CJS entries require() clean (3/3 probed).

Generated by Claude Code

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 3 package(s): @objectstack/plugin-webhooks, @objectstack/service-automation, @objectstack/service-messaging, touching 13 documentable anchor(s).

28 hand-written doc(s) name something this change touched — list omitted above 15 rows. Re-derive on the tree named below: node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0.

⛔ 5 release-owned page(s) also affected — read-only, see AGENTS.md Documentation Guardrails.

What this run could not see

• 1 cross-cutting symbol(s) contributed no route anchor: organizationId (4 routes)
• 2 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 47 of 219 client-bound route-ledger rows — the other 172 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 172: 14 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 56 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 102 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.

Coarse fallback — 6 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 214ce1a01bdf23e06de3ef9a877c3a36013531fd — the merge of head 479e6315071c3b8cef710d46da7e64b4d3a160c1 into base 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 214ce1a01bdf23e06de3ef9a877c3a36013531fd && git checkout 214ce1a01bdf23e06de3ef9a877c3a36013531fd
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 479e6315071c3b8cef710d46da7e64b4d3a160c1 && git checkout -B drift-repro 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 && git merge --no-ff 479e6315071c3b8cef710d46da7e64b4d3a160c1
node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[os-steve]

Fixes#13546

What breaks today

sys_http_delivery is tenant-scoped and redeliver() — its one request-reachable door (POST /api/v1/webhooks/redeliver, open to any authenticated user) — deliberately scopes by the caller's organization (#10740). But the enqueue door never stamped organization_id, and the SQL driver's tenant term is (organization_id = :tenantId OR organization_id IS NULL) — a deliberate global-row fail-open. So 100% of rows landed in the NULL arm: visible to, and replayable by, every organization on a walled deployment. Declared, not enforced; this PR restores the declared contract. The driver's tenant term is untouched — removing the IS NULL arm would hide true global rows from every tenant, a different defect.

The repair — mirror of the sibling that already received it

The notification outbox got exactly this fix (EnqueueDeliveryInput.organizationId, SqlOutbox.enqueue writing organization_id: input.organizationId ?? null). This PR gives the HTTP outbox the same shape, end to end:

• Contract (http-outbox.ts): EnqueueHttpInput gains an organizationId?: string member — inherited by UndeliverableHttpInput, so parked rows are tenant-stamped too — and HttpDelivery surfaces it on read-back.
• SQL impl (sql-http-outbox.ts): the shared insert() (reached from both enqueue() and recordUndeliverable()) writes organization_id: input.organizationId ?? null, the same line the sibling writes; toDelivery() maps it back (NULL → absent).
• Memory impl (memory-http-outbox.ts): stores the same field — and, now that its rows carry a tenant, its redeliver() honors the obligation its own tenant-audit, the update half: ack is a dispatcher sweep but redeliver is request-reachable — two sites on one object with OPPOSITE classifications #10740 note recorded ("a future memory implementation that DOES store a tenant owes the predicate here"), with the driver's exact semantics: another organization's row is invisible (RESOURCE_NOT_FOUND, never an existence oracle), an org-less row stays a global row any tenant may replay, a tenant-less caller stays unscoped.
• Producer 1 — flow http node (http-nodes.ts, durable mode): threads the run's acting organization, AutomationContext.tenantId — the same source and the same no-fallback rule as the notify node's [finding] sys_inbox_message/sys_notification/sys_email 等平台表从不写 organization_id(存量与新增行 100% null)——请确认多组织语义是否设计如此 #11303 repair in the same directory. When a run has none, it warns loudly (fail-loud, not fail-closed, not fail-guess) and enqueues org-less.
• Producer 2 — webhook auto-enqueuer (auto-enqueuer.ts): stamps each delivery with its subscription's own organization (sys_webhook is organization-scoped, Five more instances of the #8323 class: admin- and user-authored names on tenant-scoped objects still carry installation-wide unique indexes #8554, so the row carries the tenant that authored it). There is no request context on this fire-and-forget path; the subscription row is the one honest source. Org-less subscriptions enqueue org-less, silently — a per-event warn here would fire on every record write of a single-posture install.

Forward-stamping only. Existing NULL rows are untouched; their disposition is the separate decision the card names. No backfill, no driver changes, no schema changes (organization_id is kernel-provisioned on every object).

Optional vs required — argued, not assumed

The new member is optional (organizationId?: string). Weighed explicitly against the required-but-undefined-able shape this same file uses for RedeliverOptions.tenantId:

• For optional: it mirrors the sibling notification outbox's EnqueueDeliveryInput.organizationId and EmitInput.organizationId — one convention across the two outboxes of one package (a second convention here would be this card's defect repeated). It is non-breaking for the published surface. And org-less enqueues are legitimate: a single-posture install and every stack before its first organization must keep enqueuing.
• For required (organizationId: string | undefined): nothing tells a future producer it forgot one — the exact gap that produced this bug. That is a real cost. The mitigations shipped here: the producer population of this seam is enumerated (two, both repaired in this PR), the flow node warns loudly on an org-less multi-org run, and the driver's auditMissingTenant line still reports unscoped writes on multi-tenant boots. RedeliverOptions.tenantId went required because its callers are arbitrary request routes; this seam's callers are the two service-internal producers.

If the contract reviewer prefers the required shape, the change is mechanical (both producers already always spell the key or spread it conditionally) but it becomes a compile-breaking widening of a published input.

Write-surface enumeration (the triage's must-answer)

Is SqlHttpOutbox.insert the only door? Enumerated, not exemplified, at origin/main (ff37576):

• Production constructors of an IHttpOutbox: exactly one — messaging-service-plugin.ts:310 (new SqlHttpOutbox(engine, ...), default object name). MemoryHttpOutbox is constructed only by tests.
• Row-creating engine calls on the object: exactly one INSERT site, SqlHttpOutbox.insert, reached from enqueue() and recordUndeliverable() — both covered by this stamp. No bulkCreate/upsert/seed writes anywhere (grep -rn "insert(SYS_HTTP_DELIVERY|'sys_http_delivery'" over packages/ apps/: only the four test files that build their own rows).
• The remaining writes (claim/ack/redeliver UPDATEs) create no rows and never touch the tenant column; userActions declares create/edit/delete false and the object is managedBy: 'engine-owned'.

Verification

• New pins: packages/services/service-messaging/src/http-outbox-organization.test.ts (9 — SQL stamp verbatim / explicit-NULL normalization / parked-row stamp / read-back mapping; memory parity + all three arms of the redeliver predicate, ADR-0112 code asserted on the refusal); http-nodes.test.ts (+2 — tenant threaded verbatim; over-denial control: org-less run still enqueues, key absent, warns); auto-enqueuer.test.ts (+3 — per-record, org-less control, bulk path).
• Targeted suites at head sha e2ecb71: service-messaging 7 files / 48 tests pass (incl. the tenant-audit and parked-record integration suites), service-automation 4 files / 34 pass, plugin-webhooks 3 files / 40 pass. typecheck green for service-messaging + plugin-webhooks (service-automation declares none); --listFiles confirms the new test file is inside the messaging typecheck.
• Reverse verification (committed state, restore proven by blob hash vs HEAD): plugin-webhooks resolves @objectstack/service-messaging via dist/index.d.ts (tsc --listFiles); mutating the producer key to organizationIdX reddens typecheck with TS2561 naming organizationId as the fix — proving the rebuilt .d.ts carries the member and the literal is enforced against it. Restore verified byte-identical (worktree blob = HEAD blob 7f52692).
• Gates at e2ecb71: all 14 path-derived families green except check:dual-build-cjs-loads, which is NOT MEASURED locally (PREREQUISITE NOT MET — it reads every package's dist and 41+ untouched packages are unbuilt here; CI's Build Core owns the full run). Declared narrowing: the gate's own population is every published package's built entry (quoted from its error output); this diff touches 3 packages and no build/exports config of any other, so untouched packages' verdicts cannot move; the 3 touched packages were rebuilt and their CJS entries require() clean (3/3 probed).

Generated by Claude Code

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 3 package(s): @objectstack/plugin-webhooks, @objectstack/service-automation, @objectstack/service-messaging, touching 13 documentable anchor(s).

28 hand-written doc(s) name something this change touched — list omitted above 15 rows. Re-derive on the tree named below: node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0.

⛔ 5 release-owned page(s) also affected — read-only, see AGENTS.md Documentation Guardrails.

What this run could not see

• 1 cross-cutting symbol(s) contributed no route anchor: organizationId (4 routes)
• 2 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 47 of 219 client-bound route-ledger rows — the other 172 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 172: 14 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 56 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 102 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.

Coarse fallback — 6 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 214ce1a01bdf23e06de3ef9a877c3a36013531fd — the merge of head 479e6315071c3b8cef710d46da7e64b4d3a160c1 into base 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 214ce1a01bdf23e06de3ef9a877c3a36013531fd && git checkout 214ce1a01bdf23e06de3ef9a877c3a36013531fd
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 479e6315071c3b8cef710d46da7e64b4d3a160c1 && git checkout -B drift-repro 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 && git merge --no-ff 479e6315071c3b8cef710d46da7e64b4d3a160c1
node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[os-steve]

Fixes#13546

What breaks today

sys_http_delivery is tenant-scoped and redeliver() — its one request-reachable door (POST /api/v1/webhooks/redeliver, open to any authenticated user) — deliberately scopes by the caller's organization (#10740). But the enqueue door never stamped organization_id, and the SQL driver's tenant term is (organization_id = :tenantId OR organization_id IS NULL) — a deliberate global-row fail-open. So 100% of rows landed in the NULL arm: visible to, and replayable by, every organization on a walled deployment. Declared, not enforced; this PR restores the declared contract. The driver's tenant term is untouched — removing the IS NULL arm would hide true global rows from every tenant, a different defect.

The repair — mirror of the sibling that already received it

The notification outbox got exactly this fix (EnqueueDeliveryInput.organizationId, SqlOutbox.enqueue writing organization_id: input.organizationId ?? null). This PR gives the HTTP outbox the same shape, end to end:

• Contract (http-outbox.ts): EnqueueHttpInput gains an organizationId?: string member — inherited by UndeliverableHttpInput, so parked rows are tenant-stamped too — and HttpDelivery surfaces it on read-back.
• SQL impl (sql-http-outbox.ts): the shared insert() (reached from both enqueue() and recordUndeliverable()) writes organization_id: input.organizationId ?? null, the same line the sibling writes; toDelivery() maps it back (NULL → absent).
• Memory impl (memory-http-outbox.ts): stores the same field — and, now that its rows carry a tenant, its redeliver() honors the obligation its own tenant-audit, the update half: ack is a dispatcher sweep but redeliver is request-reachable — two sites on one object with OPPOSITE classifications #10740 note recorded ("a future memory implementation that DOES store a tenant owes the predicate here"), with the driver's exact semantics: another organization's row is invisible (RESOURCE_NOT_FOUND, never an existence oracle), an org-less row stays a global row any tenant may replay, a tenant-less caller stays unscoped.
• Producer 1 — flow http node (http-nodes.ts, durable mode): threads the run's acting organization, AutomationContext.tenantId — the same source and the same no-fallback rule as the notify node's [finding] sys_inbox_message/sys_notification/sys_email 等平台表从不写 organization_id(存量与新增行 100% null)——请确认多组织语义是否设计如此 #11303 repair in the same directory. When a run has none, it warns loudly (fail-loud, not fail-closed, not fail-guess) and enqueues org-less.
• Producer 2 — webhook auto-enqueuer (auto-enqueuer.ts): stamps each delivery with its subscription's own organization (sys_webhook is organization-scoped, Five more instances of the #8323 class: admin- and user-authored names on tenant-scoped objects still carry installation-wide unique indexes #8554, so the row carries the tenant that authored it). There is no request context on this fire-and-forget path; the subscription row is the one honest source. Org-less subscriptions enqueue org-less, silently — a per-event warn here would fire on every record write of a single-posture install.

Forward-stamping only. Existing NULL rows are untouched; their disposition is the separate decision the card names. No backfill, no driver changes, no schema changes (organization_id is kernel-provisioned on every object).

Optional vs required — argued, not assumed

The new member is optional (organizationId?: string). Weighed explicitly against the required-but-undefined-able shape this same file uses for RedeliverOptions.tenantId:

• For optional: it mirrors the sibling notification outbox's EnqueueDeliveryInput.organizationId and EmitInput.organizationId — one convention across the two outboxes of one package (a second convention here would be this card's defect repeated). It is non-breaking for the published surface. And org-less enqueues are legitimate: a single-posture install and every stack before its first organization must keep enqueuing.
• For required (organizationId: string | undefined): nothing tells a future producer it forgot one — the exact gap that produced this bug. That is a real cost. The mitigations shipped here: the producer population of this seam is enumerated (two, both repaired in this PR), the flow node warns loudly on an org-less multi-org run, and the driver's auditMissingTenant line still reports unscoped writes on multi-tenant boots. RedeliverOptions.tenantId went required because its callers are arbitrary request routes; this seam's callers are the two service-internal producers.

If the contract reviewer prefers the required shape, the change is mechanical (both producers already always spell the key or spread it conditionally) but it becomes a compile-breaking widening of a published input.

Write-surface enumeration (the triage's must-answer)

Is SqlHttpOutbox.insert the only door? Enumerated, not exemplified, at origin/main (ff37576):

• Production constructors of an IHttpOutbox: exactly one — messaging-service-plugin.ts:310 (new SqlHttpOutbox(engine, ...), default object name). MemoryHttpOutbox is constructed only by tests.
• Row-creating engine calls on the object: exactly one INSERT site, SqlHttpOutbox.insert, reached from enqueue() and recordUndeliverable() — both covered by this stamp. No bulkCreate/upsert/seed writes anywhere (grep -rn "insert(SYS_HTTP_DELIVERY|'sys_http_delivery'" over packages/ apps/: only the four test files that build their own rows).
• The remaining writes (claim/ack/redeliver UPDATEs) create no rows and never touch the tenant column; userActions declares create/edit/delete false and the object is managedBy: 'engine-owned'.

Verification

• New pins: packages/services/service-messaging/src/http-outbox-organization.test.ts (9 — SQL stamp verbatim / explicit-NULL normalization / parked-row stamp / read-back mapping; memory parity + all three arms of the redeliver predicate, ADR-0112 code asserted on the refusal); http-nodes.test.ts (+2 — tenant threaded verbatim; over-denial control: org-less run still enqueues, key absent, warns); auto-enqueuer.test.ts (+3 — per-record, org-less control, bulk path).
• Targeted suites at head sha e2ecb71: service-messaging 7 files / 48 tests pass (incl. the tenant-audit and parked-record integration suites), service-automation 4 files / 34 pass, plugin-webhooks 3 files / 40 pass. typecheck green for service-messaging + plugin-webhooks (service-automation declares none); --listFiles confirms the new test file is inside the messaging typecheck.
• Reverse verification (committed state, restore proven by blob hash vs HEAD): plugin-webhooks resolves @objectstack/service-messaging via dist/index.d.ts (tsc --listFiles); mutating the producer key to organizationIdX reddens typecheck with TS2561 naming organizationId as the fix — proving the rebuilt .d.ts carries the member and the literal is enforced against it. Restore verified byte-identical (worktree blob = HEAD blob 7f52692).
• Gates at e2ecb71: all 14 path-derived families green except check:dual-build-cjs-loads, which is NOT MEASURED locally (PREREQUISITE NOT MET — it reads every package's dist and 41+ untouched packages are unbuilt here; CI's Build Core owns the full run). Declared narrowing: the gate's own population is every published package's built entry (quoted from its error output); this diff touches 3 packages and no build/exports config of any other, so untouched packages' verdicts cannot move; the 3 touched packages were rebuilt and their CJS entries require() clean (3/3 probed).

Generated by Claude Code

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 3 package(s): @objectstack/plugin-webhooks, @objectstack/service-automation, @objectstack/service-messaging, touching 13 documentable anchor(s).

28 hand-written doc(s) name something this change touched — list omitted above 15 rows. Re-derive on the tree named below: node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0.

⛔ 5 release-owned page(s) also affected — read-only, see AGENTS.md Documentation Guardrails.

What this run could not see

• 1 cross-cutting symbol(s) contributed no route anchor: organizationId (4 routes)
• 2 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 47 of 219 client-bound route-ledger rows — the other 172 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 172: 14 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 56 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 102 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.

Coarse fallback — 6 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 214ce1a01bdf23e06de3ef9a877c3a36013531fd — the merge of head 479e6315071c3b8cef710d46da7e64b4d3a160c1 into base 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 214ce1a01bdf23e06de3ef9a877c3a36013531fd && git checkout 214ce1a01bdf23e06de3ef9a877c3a36013531fd
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 479e6315071c3b8cef710d46da7e64b4d3a160c1 && git checkout -B drift-repro 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 && git merge --no-ff 479e6315071c3b8cef710d46da7e64b4d3a160c1
node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[os-steve]

Fixes#13546

What breaks today

sys_http_delivery is tenant-scoped and redeliver() — its one request-reachable door (POST /api/v1/webhooks/redeliver, open to any authenticated user) — deliberately scopes by the caller's organization (#10740). But the enqueue door never stamped organization_id, and the SQL driver's tenant term is (organization_id = :tenantId OR organization_id IS NULL) — a deliberate global-row fail-open. So 100% of rows landed in the NULL arm: visible to, and replayable by, every organization on a walled deployment. Declared, not enforced; this PR restores the declared contract. The driver's tenant term is untouched — removing the IS NULL arm would hide true global rows from every tenant, a different defect.

The repair — mirror of the sibling that already received it

The notification outbox got exactly this fix (EnqueueDeliveryInput.organizationId, SqlOutbox.enqueue writing organization_id: input.organizationId ?? null). This PR gives the HTTP outbox the same shape, end to end:

• Contract (http-outbox.ts): EnqueueHttpInput gains an organizationId?: string member — inherited by UndeliverableHttpInput, so parked rows are tenant-stamped too — and HttpDelivery surfaces it on read-back.
• SQL impl (sql-http-outbox.ts): the shared insert() (reached from both enqueue() and recordUndeliverable()) writes organization_id: input.organizationId ?? null, the same line the sibling writes; toDelivery() maps it back (NULL → absent).
• Memory impl (memory-http-outbox.ts): stores the same field — and, now that its rows carry a tenant, its redeliver() honors the obligation its own tenant-audit, the update half: ack is a dispatcher sweep but redeliver is request-reachable — two sites on one object with OPPOSITE classifications #10740 note recorded ("a future memory implementation that DOES store a tenant owes the predicate here"), with the driver's exact semantics: another organization's row is invisible (RESOURCE_NOT_FOUND, never an existence oracle), an org-less row stays a global row any tenant may replay, a tenant-less caller stays unscoped.
• Producer 1 — flow http node (http-nodes.ts, durable mode): threads the run's acting organization, AutomationContext.tenantId — the same source and the same no-fallback rule as the notify node's [finding] sys_inbox_message/sys_notification/sys_email 等平台表从不写 organization_id(存量与新增行 100% null)——请确认多组织语义是否设计如此 #11303 repair in the same directory. When a run has none, it warns loudly (fail-loud, not fail-closed, not fail-guess) and enqueues org-less.
• Producer 2 — webhook auto-enqueuer (auto-enqueuer.ts): stamps each delivery with its subscription's own organization (sys_webhook is organization-scoped, Five more instances of the #8323 class: admin- and user-authored names on tenant-scoped objects still carry installation-wide unique indexes #8554, so the row carries the tenant that authored it). There is no request context on this fire-and-forget path; the subscription row is the one honest source. Org-less subscriptions enqueue org-less, silently — a per-event warn here would fire on every record write of a single-posture install.

Forward-stamping only. Existing NULL rows are untouched; their disposition is the separate decision the card names. No backfill, no driver changes, no schema changes (organization_id is kernel-provisioned on every object).

Optional vs required — argued, not assumed

The new member is optional (organizationId?: string). Weighed explicitly against the required-but-undefined-able shape this same file uses for RedeliverOptions.tenantId:

• For optional: it mirrors the sibling notification outbox's EnqueueDeliveryInput.organizationId and EmitInput.organizationId — one convention across the two outboxes of one package (a second convention here would be this card's defect repeated). It is non-breaking for the published surface. And org-less enqueues are legitimate: a single-posture install and every stack before its first organization must keep enqueuing.
• For required (organizationId: string | undefined): nothing tells a future producer it forgot one — the exact gap that produced this bug. That is a real cost. The mitigations shipped here: the producer population of this seam is enumerated (two, both repaired in this PR), the flow node warns loudly on an org-less multi-org run, and the driver's auditMissingTenant line still reports unscoped writes on multi-tenant boots. RedeliverOptions.tenantId went required because its callers are arbitrary request routes; this seam's callers are the two service-internal producers.

If the contract reviewer prefers the required shape, the change is mechanical (both producers already always spell the key or spread it conditionally) but it becomes a compile-breaking widening of a published input.

Write-surface enumeration (the triage's must-answer)

Is SqlHttpOutbox.insert the only door? Enumerated, not exemplified, at origin/main (ff37576):

• Production constructors of an IHttpOutbox: exactly one — messaging-service-plugin.ts:310 (new SqlHttpOutbox(engine, ...), default object name). MemoryHttpOutbox is constructed only by tests.
• Row-creating engine calls on the object: exactly one INSERT site, SqlHttpOutbox.insert, reached from enqueue() and recordUndeliverable() — both covered by this stamp. No bulkCreate/upsert/seed writes anywhere (grep -rn "insert(SYS_HTTP_DELIVERY|'sys_http_delivery'" over packages/ apps/: only the four test files that build their own rows).
• The remaining writes (claim/ack/redeliver UPDATEs) create no rows and never touch the tenant column; userActions declares create/edit/delete false and the object is managedBy: 'engine-owned'.

Verification

• New pins: packages/services/service-messaging/src/http-outbox-organization.test.ts (9 — SQL stamp verbatim / explicit-NULL normalization / parked-row stamp / read-back mapping; memory parity + all three arms of the redeliver predicate, ADR-0112 code asserted on the refusal); http-nodes.test.ts (+2 — tenant threaded verbatim; over-denial control: org-less run still enqueues, key absent, warns); auto-enqueuer.test.ts (+3 — per-record, org-less control, bulk path).
• Targeted suites at head sha e2ecb71: service-messaging 7 files / 48 tests pass (incl. the tenant-audit and parked-record integration suites), service-automation 4 files / 34 pass, plugin-webhooks 3 files / 40 pass. typecheck green for service-messaging + plugin-webhooks (service-automation declares none); --listFiles confirms the new test file is inside the messaging typecheck.
• Reverse verification (committed state, restore proven by blob hash vs HEAD): plugin-webhooks resolves @objectstack/service-messaging via dist/index.d.ts (tsc --listFiles); mutating the producer key to organizationIdX reddens typecheck with TS2561 naming organizationId as the fix — proving the rebuilt .d.ts carries the member and the literal is enforced against it. Restore verified byte-identical (worktree blob = HEAD blob 7f52692).
• Gates at e2ecb71: all 14 path-derived families green except check:dual-build-cjs-loads, which is NOT MEASURED locally (PREREQUISITE NOT MET — it reads every package's dist and 41+ untouched packages are unbuilt here; CI's Build Core owns the full run). Declared narrowing: the gate's own population is every published package's built entry (quoted from its error output); this diff touches 3 packages and no build/exports config of any other, so untouched packages' verdicts cannot move; the 3 touched packages were rebuilt and their CJS entries require() clean (3/3 probed).

Generated by Claude Code

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

This PR changes 3 package(s): @objectstack/plugin-webhooks, @objectstack/service-automation, @objectstack/service-messaging, touching 13 documentable anchor(s).

28 hand-written doc(s) name something this change touched — list omitted above 15 rows. Re-derive on the tree named below: node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0.

⛔ 5 release-owned page(s) also affected — read-only, see AGENTS.md Documentation Guardrails.

What this run could not see

• 1 cross-cutting symbol(s) contributed no route anchor: organizationId (4 routes)
• 2 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 47 of 219 client-bound route-ledger rows — the other 172 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 172: 14 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 56 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 102 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.

Coarse fallback — 6 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 214ce1a01bdf23e06de3ef9a877c3a36013531fd — the merge of head 479e6315071c3b8cef710d46da7e64b4d3a160c1 into base 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 214ce1a01bdf23e06de3ef9a877c3a36013531fd && git checkout 214ce1a01bdf23e06de3ef9a877c3a36013531fd
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 479e6315071c3b8cef710d46da7e64b4d3a160c1 && git checkout -B drift-repro 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 && git merge --no-ff 479e6315071c3b8cef710d46da7e64b4d3a160c1
node scripts/docs-audit/affected-docs.mjs --json 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 51ecb2fdd2035a8924603b497a6b7a2ee8277fb0 → pass the list as
args.docs, on the commit named under Which tree this was computed on.