feat(besluitvorming): a voorstel is a case, not a record of its own

[rubenvdlinde]

Why

The dossiq menu carried a Proposals entry nobody could account for. Following it back changed what the fix should be.

A voorstel is not a thing in its own right. It is a case whose case type requires a decision. It was modelled a second time as a standalone object, with its own list page, its own status vocabulary and its own sign-off engine.

Nothing moved it here from decidiq, which was the working assumption. voorstel has been in this app since its earliest schema work, under the old procest name, long before decidiq existed as a target. docs/decisions/besluitvorming-vs-decidesk.md is the reason it stayed: on 2026-06-22 it recommended keeping besluitvorming here and explicitly not folding it into decidesk. Five later changes moved the other way and nobody revoked the doc, so it kept telling readers the opposite of what the fleet was doing. It is marked superseded here.

The menu entry was a gate artefact, not a decision. src/menu-layout.json says so: Voorstellen had been removed, and was restored on 2026-09-02 because the case-detail consolidation dropped the case-voorstellen widget, the only navigation edge gate-53 could follow to /voorstellen. That note asked whichever change retired the surface for real to re-remove the entry with a truthful waiver.

What replaces it

The abstraction was already shipped, in three pieces.

The voorstel surface did	It is done here now
list the proposals	Cases, narrowed by case type through folderSidebar.filterField
show one proposal and its chain	CaseDetail, and its case-decidesk-decisions leaf
say which decision is needed	caseType.decisionTypes
run the sign-off chain	decidiq's ApprovalRoute engine
record the outcome on the case	case.decisions

The three besluitvorming case types (College-besluit, Raadsbesluit, Mandaatbesluit) are that abstraction, and they survive untouched. They keep their Voorstel opstellen status, their Steller role and their voorstel document types, because that word is the case type's own vocabulary. What retires is the pretence that a proposal is a record you file somewhere else.

No gate-53 waiver, and none is owed

The page is deleted rather than hidden, so there is no orphaned route for a removals entry to speak for. That answers the question the 2026-09-02 note left open, and the answer is that the question did not apply. gate-53 reports 7 warnings, all pre-existing, none naming Voorstellen.

Two defects this found

Neither is the point of the change, and both say something about how the surface was maintained.

A row action that had never worked. voorstelReminder POSTed to /apps/dossiq/api/notifications/parafering-reminder. That URL appears nowhere in appinfo/routes.php, at any commit. Every click answered 404 into an empty catch that logged to the console.

A guard that would have stranded every besluitvorming case. The three bundles declared a voorstelStatus guard. An unknown guard type evaluates as passed: false in GuardRegistry, so deleting the guard class alone would have locked every besluitvorming case at Parafering on every install, with "Onbekende guard" as the only evidence. The bundles are amended in the same commit.

A third one, in this PR's own first commit

The retirement edited docs/features.json by hand. That file is generated: the pre-commit hook rebuilds it from openspec/features.overlay.json and silently put the entry back inside the same commit that removed it, printing one line and exiting 0. The second commit fixes it at the source. Worth knowing about before the next person edits that file.

What is deliberately not done

Stored rows are not deleted. Retiring a schema does not remove what an install already holds. RenameDutchColumns and RenameDutchSchemaSlugs keep their voorstel entries: they are historical migrations that run over exactly those rows.

Five sibling changes are removed rather than left standing. parafering-to-decidiq, parafering-runtime-to-decidiq, parafering-runs-as-a-flow and approval-routes-are-flows all existed to move the engine somewhere; deleting it reaches the same end state and asks nothing of decidiq. bw-voorstellen-view existed to build the list view /voorstellen never had, which is the one clear waste here.

StubApiDriftTest is left alone. It compares tests/Stubs/ against a sibling checkout at ../openregister, which in a shared dev workspace is another session's live working tree. It was red on clean development, and then reported drift in the opposite direction twenty minutes later because that checkout switched branch. CI clones openregister fresh, so its verdict is the one that counts.

Verification

Check	Result
PHPUnit	2911 tests, 0 failures, 0 warnings, 1 pre-existing skip
Vitest	35 files, 359 tests, all green
ESLint	0 errors, warnings 839 to 823
PHPCS	0 errors
PHPMD	clean on both rulesets
Psalm	no errors
PHPStan	no errors
Hydra gates	all 82 applicable green, with --base origin/development

PHPStan caught a real leftover on the way: $bundle became an unused closure binding once the parafeerroute read went.

Review this first

src/menu-layout.json and openspec/changes/proposals-are-cases/proposal.md carry the reasoning. The 17,441 deleted lines are mostly the four schemas and their seed data.

🤖 Generated with Claude Code

[github-actions]

Quality Report — ConductionNL/dossiq @ 1960d16

Check	PHP	Vue	Security	License	Tests
lint	✅
phpcs	✅
phpmd	✅
psalm	✅
phpstan	✅
phpmetrics	✅
eslint		❌
stylelint		✅
build		✅
check-manifest		✅
check-vue3-compile		✅
test-l10n		✅
format		❌
check-schema-l10n		✅
check-l10n-js		✅
composer			✅	✅ 106/106
npm			✅	✅ 540/540
app:check-code	⏭️
info.xml	✅
REUSE	❌
PHPUnit					❌
Newman					⏭️
Playwright					⏭️ deferred — runs on the promotion into beta/main, not on a pull request into development
Hydra gates					✅

---

Quality workflow — 2026-09-03 20:31 UTC

Download the full PDF report from the workflow artifacts.

[github-actions]

Quality Report — ConductionNL/dossiq @ bdcd34b

Check	PHP	Vue	Security	License	Tests
lint	✅
phpcs	✅
phpmd	✅
psalm	✅
phpstan	✅
phpmetrics	✅
eslint		✅
stylelint		✅
build		✅
check-manifest		✅
check-vue3-compile		✅
test-l10n		✅
format		✅
check-schema-l10n		✅
check-l10n-js		✅
composer			✅	✅ 106/106
npm			✅	✅ 540/540
app:check-code	⏭️
info.xml	✅
REUSE	❌
PHPUnit					❌
Newman					⏭️
Playwright					⏭️ deferred — runs on the promotion into beta/main, not on a pull request into development
Hydra gates					✅

---

Quality workflow — 2026-09-03 21:23 UTC

Download the full PDF report from the workflow artifacts.

[github-actions]

Quality Report — ConductionNL/dossiq @ 993c372

Check	PHP	Vue	Security	License	Tests
lint	✅
phpcs	✅
phpmd	✅
psalm	✅
phpstan	✅
phpmetrics	✅
eslint		✅
stylelint		✅
build		✅
check-manifest		✅
check-vue3-compile		✅
test-l10n		✅
format		✅
check-schema-l10n		✅
check-l10n-js		✅
composer			✅	✅ 106/106
npm			✅	✅ 540/540
app:check-code	⏭️
info.xml	✅
REUSE	❌
PHPUnit					✅
Newman					⏭️
Playwright					⏭️ deferred — runs on the promotion into beta/main, not on a pull request into development
Hydra gates					✅

---

Quality workflow — 2026-09-03 21:34 UTC

Download the full PDF report from the workflow artifacts.

[rubenvdlinde]

Superseded by #1796, which carries the identical tree (same tree hash) with the Co-Authored-By trailers stripped from the six commits authored here.

Closed rather than updated because rewriting history needs a force-push, which this workspace's block-fleet-force-push guard denies, and narrowing that guard is not a change to make on an agent's own judgement.