Skip to content

chore(openspec): finish the two changes the triage could not decide - #2066

Merged
rubenvdlinde merged 2 commits into
developmentfrom
chore/finish-two-openspec-changes
Sep 9, 2026
Merged

rubenvdlinde merged 2 commits into
developmentfrom
chore/finish-two-openspec-changes

Conversation

@rubenvdlinde

Copy link
Copy Markdown
Contributor

What this is

The two changes #2057 could not finish. openspec/changes/ goes from 22 to 20.

Both were flagged there as undecidable, and both turned out to be answerable. The website half of
beta-surface-alignment is live in conduction-website, which makes it a small remnant rather than a
change that never landed, and subsidieregeling-is-a-casetype needed a delta written rather than a
directory deleted.

1. beta-surface-alignment: the fourth surface

Three of its four ticked tasks were real. conduction-website carries
src/pages/apps/dossiq.mdx and i18n/nl/docusaurus-plugin-content-pages/apps/dossiq.mdx on main.
The fourth, appinfo/info.xml, was ticked in July with the edit absent: the description still
undersold the app as generic case management in ten bullets, and still claimed the app is "modelled on
CMMN 1.1 and the ZGW APIs".

Seven bullets are added, EN and NL in parallel: VTH process templates, objections and appeals, Woo
requests, workflow board, deadline monitoring with dwangsom, case map, appointment booking. They sit
under a second heading rather than growing one list to seventeen items, because an App Store
description is read rather than counted.

The standards line now reads deep ZGW API alignment, with workflow templates modelled on CMMN
case-management concepts. There is no CMMN runtime anywhere in lib/, and docs/Features/README.md
already marked CMMN 1.1 as partial.

Every bullet was re-verified against the tree, not against the July proposal. That was not
ceremony. The navigation has moved a great deal since July, and the proposal describes an app that no
longer exists in places: bezwaar and beroep have no top-level menu entry any more and are reached as
case types, and IngebrekestellingController, which the proposal names, is gone. What is actually
there, and what each bullet is traceable to:

Bullet Backed by
VTH process templates six files in lib/Settings/seed/vth-workflow-templates/ (handhavingszaak, omgevingsvergunning regulier and uitgebreid, sloopmelding, toezichtzaak bouw and milieu)
Objections and appeals BezwaarHearingController.php, BezwaarDetail / BeroepDetail / BezwaarAdviceRequestDetail pages, beroep schema at dossiq_register.json:4671
Wet open overheid (Woo) requests WOOAssessmentController.php, lib/Settings/templates/woo-verzoek.json
Workflow board src/views/workflow-board/WorkflowBoard.vue, drag and the keyboard menu both
Deadline monitoring, dwangsom TermijnDashboard page, DwangsomController.php, DwangsomPaymentCallbackController.php
Case map CaseMap page, CasesOnMapView.vue, WmsLayersMenu
Appointment booking AppointmentController with timeslots, cancel and no-show routes at appinfo/routes.php:438-447, plus the PublicAppointment page

That is the spec's own contract: beta-alignment REQ says any feature named in info.xml has to be
traceable to a controller, service, background job or Vue view that ships.

The archived tasks.md records that the box was ticked in July with the work absent, so the archive
tells the truth about what happened rather than reading as if it always had.

2. subsidieregeling-is-a-casetype: the delta it never had

The code shipped on 2026-08-27 in 05919d25 and the directory then sat in openspec/changes/ for two
weeks, because it was a proposal and nothing else. Not deleted: a shipped capability with no spec is
an afternoon somebody spends re-deriving it later.

Written on case-types, not subsidieverlening-keten. Every requirement here is about how a
blueprint is modelled and administered, not about the grant chain's lifecycle. The four field mappings,
the propertyType extension, the propertyDefinition records and the index the schemes moved onto are
all case-type machinery; nothing touches aanvraag, beoordeling, beschikking or vaststelling. There is a
second reason too: subsidieverlening-keten carries status: partial with a careful note tracking
which of its requirements do not actually run, and dropping three that do run into it would blur a
distinction somebody worked to make.

  • REQ-CT-21 a grant scheme is a case type. Includes the integer to ISO-8601 conversion of the
    decision term, because an integer stores happily and is understood by neither the renderer nor the
    Awb 4:13 deadline maths.
  • REQ-CT-22 a property definition can carry an enum or a JSON document. This is the whole reason
    propertyType gained those two values: flattening a four-value enum to a string keeps the value and
    loses the constraint, and nothing raises.
  • REQ-CT-23 schemes are administered on the Case types index, the retired /subsidieregelingen
    route falls through rather than erroring, and subsidieRegeling is kept one release on purpose so
    the migration can still read the rows it converts.

All six scenarios carry @e2e tests/e2e/spec-coverage/subsidieregeling-is-a-casetype.spec.ts, which is
where they were already asserted, six tests across two describes. tasks.md is written retroactively
with the evidence on each box rather than a bare tick, which is the point of the whole exercise.

One box stays open deliberately: deleting the deprecated subsidieRegeling schema, which waits until
the migration has run everywhere. Deleting it now would leave the migration reading zero rows and
reporting the same success as one that had work to do.

The six anchors this PR did fix

The e2e file's six // @e2e comments pointed at openspec/changes/subsidieregeling-is-a-casetype/proposal.md,
a path this PR moves. They now name openspec/specs/case-types/spec.md. That is the six lines this
archive itself made stale. The wider anchor debt (420 anchors naming a change archived months ago, ~150
change names cited that way) is untouched and stays its own change.

Verification

openspec validate --strict --all: nothing made invalid, and subsidieregeling-is-a-casetype went
from invalid to valid before archiving. node tests/l10n/check-l10n.js exits 0. appinfo/info.xml
parses, and carries zero em-dashes and zero double-dashes in either description, so gate 96 is clean.
The tests/e2e/...spec.ts change is comment text only, same leading tab, same line count.

Ruben van der Linde added 2 commits September 9, 2026 11:24
…line stops overclaiming

`beta-surface-alignment` was ticked 16 of 16 in July with this task's work
absent. Three of its four ticked tasks were real: the product page and its Dutch
copy are live in conduction-website. The fourth, `appinfo/info.xml`, was not.
The description still undersold the app as generic case management with ten
bullets, and still read "modelled on CMMN 1.1 and the ZGW APIs".

Seven bullets added, EN and NL: VTH process templates, objections and appeals,
Woo requests, workflow board, deadline monitoring with dwangsom, case map, and
appointment booking. They sit under a second heading rather than extending one
list to seventeen items, because an App Store description is read, not counted.

The standards line now reads deep ZGW API alignment, with workflow templates
modelled on CMMN case-management concepts. There is no CMMN runtime in `lib/`,
and the app's own docs already marked CMMN 1.1 as partial.

Every bullet was re-verified against the tree, not against the July proposal.
That mattered: the navigation has moved a great deal since, so the proposal's
wording could not be trusted as a description of what ships now. Bezwaar and
beroep have no top-level menu entry any more and are reached as case types;
`IngebrekestellingController` is gone. What is there: six VTH templates in
`lib/Settings/seed/vth-workflow-templates/`, `BezwaarHearingController` with the
committee and advice-request pages, `WOOAssessmentController` with
`templates/woo-verzoek.json`, `WorkflowBoard.vue`, `TermijnDashboard` with
`DwangsomController` and `DwangsomPaymentCallbackController`, `CasesOnMapView`
with configurable WMS layers, and `AppointmentController` with timeslot, cancel
and no-show routes plus a public page.

No em-dashes, gate 96 clean, XML well-formed. The archived tasks.md records that
the box was ticked before the work landed, so the archive tells the truth.
…er had

`subsidieregeling-is-a-casetype` shipped on 2026-08-27 in 05919d2 and then sat
in `openspec/changes/` for two weeks, because the change directory was a
proposal and nothing else: no tasks, no delta, so nothing could archive it. The
code is not in doubt. The record of it was missing.

Written retroactively: a tasks.md whose every box carries the evidence it was
ticked against, and a spec delta of three requirements.

The delta goes on `case-types`, not `subsidieverlening-keten`. Every requirement
here is about how a blueprint is modelled and administered rather than about the
grant chain's lifecycle. The four field mappings (`schemeName` to `title`,
`termStart`/`termEnd` to `validFrom`/`validUntil`, `requestTermWeeks` to
`processingDeadline`, `legalBasis` to `purpose`), the `propertyType` extension,
the `propertyDefinition` records and the index the schemes moved onto are all
case-type machinery. Nothing touches aanvraag, beoordeling, beschikking or
vaststelling. `subsidieverlening-keten` also carries `status: partial` with a
careful note about which requirements do not run, and adding three that do would
muddy it.

REQ-CT-21 a grant scheme is a case type, including the integer-to-ISO-8601
conversion of the decision term, because an integer stores happily and is
understood by neither the renderer nor the Awb 4:13 maths. REQ-CT-22 a property
definition can carry an enum or a JSON document, which is the whole reason
`propertyType` gained those two: flattening a four-value enum to a string keeps
the value and loses the constraint, silently. REQ-CT-23 schemes are administered
on the Case types index, the retired route falls through rather than erroring,
and `subsidieRegeling` is kept one release on purpose so the migration can still
read what it converts.

All six scenarios carry an `@e2e` tag at
`tests/e2e/spec-coverage/subsidieregeling-is-a-casetype.spec.ts`, which is where
they were already asserted. The six reverse anchors in that file pointed at the
change proposal, a path this commit moves, so they now name
`openspec/specs/case-types/spec.md`. That is the six lines this archive made
stale, not the wider anchor sweep, which stays its own change.

One box is left open on purpose: deleting the deprecated `subsidieRegeling`
schema, which waits until the migration has run everywhere.
@rubenvdlinde
rubenvdlinde merged commit abb59be into development Sep 9, 2026
47 of 48 checks passed
@rubenvdlinde
rubenvdlinde deleted the chore/finish-two-openspec-changes branch September 9, 2026 10:00
@github-actions

github-actions Bot commented Sep 9, 2026

Copy link
Copy Markdown
Contributor

Quality Report — ConductionNL/dossiq @ b872000

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 ✅ ✅ 549/549
app:check-code ⏭️
info.xml ✅
REUSE ❌
PHPUnit ✅
Newman ⏭️
Playwright ❌
Hydra gates ✅

Quality workflow — 2026-09-09 10:01 UTC

Download the full PDF report from the workflow artifacts.

@rubenvdlinde

Copy link
Copy Markdown
Contributor Author

The two red checks are the same pre-existing pair, and the new work is green

quality / E2E Tests (Playwright) and quality / Quality Report fail. Same two as #2057, same seven
specs, and the tally reads 299 passed, 7 failed, 1 flaky.

The seven are identical to #2057's set, which was verified there as a strict subset of
development's own failures at 719b62e:

case-actions-menu.spec.ts:321        case-type-authoring-extras.spec.ts:494
case-header.spec.ts:297              case-type-authoring-extras.spec.ts:645
case-identity.spec.ts:314            contacts-domain.spec.ts:362
case-timeline.spec.ts:98

One correction worth making explicitly, because a naive grep ✘ gets it wrong.
case-list-lenses.spec.ts:444 appears with a ✘ and looks like an eighth failure. It is the run's
one flaky: attempt 1 failed at 30.2s, retry #1 passed at 18.8s. Counting first attempts would have
reported eight failures where the runner reports seven. The set above is computed by keeping only
specs with no passing attempt at all, which is what reconciles with the tally.

This branch touches nothing those seven exercise. The whole diff is appinfo/info.xml, nine files
under openspec/, and six comment lines in one e2e file:

appinfo/info.xml
openspec/changes/archive/2026-09-09-beta-surface-alignment/*        (4 files, moved)
openspec/changes/archive/2026-09-09-subsidieregeling-is-a-casetype/* (3 files)
openspec/specs/beta-alignment/spec.md
openspec/specs/case-types/spec.md
tests/e2e/spec-coverage/subsidieregeling-is-a-casetype.spec.ts       (comments only)

The spec written here is backed by six green tests

All six tests behind REQ-CT-21 to REQ-CT-23 passed on this run, which is the point: the delta was
written to describe behaviour that runs, not behaviour that was proposed.

Test Result Requirement
both schemes exist as case types, with their fields carried ✓ 709ms REQ-CT-21
requestTermWeeks became an ISO-8601 duration, not a bare integer ✓ 557ms REQ-CT-21
the grant-specific properties came across as property definitions ✓ 1.8s REQ-CT-21
the enum property kept its allowed values instead of flattening ✓ 1.0s REQ-CT-22
the Subsidy schemes menu entry is gone ✓ 18.2s REQ-CT-23
the retired route no longer renders a scheme index ✓ 29.9s REQ-CT-23

The other 40 checks pass, including l10n coverage, Frontend Check (check:manifest) and the hydra
gates that read appinfo/info.xml. The 7 skips are the usual conditional jobs.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant