Skip to content

Define control applicability, provider dependencies, and effect decisions #1352

Closed

Description

@Brad-Edwards

The admitted apparatus is conflated with the providers applicable to one crossing. The provider protocol cannot carry predecessor results, and effect phase handling can leave a denied parent eligible.

Acceptance criteria

  • Distinguish admitted apparatus, required profile obligations, the applicable provider subset, proven inapplicability, and unresolved coverage at a specific state cut.
  • Define dependency evaluation, predecessor-result inputs, shared-state scope, composition order, and atomic commit. Missing mandatory results cannot be treated as an empty successful evaluation.
  • Define parent decisions, effect targets, execution prerequisites, independent consequences, and phase ordering. A deny/withhold decision must have the same meaning in every phase.
  • Define mandatory dependency closure and optional-provider failure. Optional advice cannot acquire veto authority; bounded support is usable only when its admitted strength satisfies the requirement.
  • Publish the decision, SEM-235/API-424 requirement amendments, ADR-108 clarifications, worked cases, and contract migration rules. Preserve separate admission, authority, and realization decisions.

References: Diagnosis, Applicability counterexample, Modular semantics.

Requirements

  • SEM-235 — Modular Participant Control and Extensible Dynamic IFC Semantics
  • API-424 — Participant-Control Provider, Composition and Effect Contracts

Activity

  1. added this to the Runtime Refactor milestone on Sep 21, 2026
  2. Brad-Edwards commented on Sep 22, 2026

    @Brad-Edwards
    CollaboratorAuthor

    🛠️ Picked up by /implement - driver codex, branch 1352-control-applicability-decisions, 2026-09-22T05:18:45.381Z.

  3. Brad-Edwards commented on Sep 22, 2026

    @Brad-Edwards
    CollaboratorAuthor

    gc workflow phase recorded: preflight (issue #1352). Posted by the MCP server to enforce ordering between workflow steps (issue #794 MVP-2). Do not edit or delete — used by downstream tools to gate phase prerequisites.

  4. Brad-Edwards commented on Sep 22, 2026

    @Brad-Edwards
    CollaboratorAuthor

    gc workflow phase recorded: preflight (issue #1352). Posted by the MCP server to enforce ordering between workflow steps (issue #794 MVP-2). Do not edit or delete — used by downstream tools to gate phase prerequisites.

  5. Brad-Edwards commented on Sep 22, 2026

    @Brad-Edwards
    CollaboratorAuthor

    gc workflow phase recorded: plan (issue #1352). Posted by the MCP server to enforce ordering between workflow steps (issue #794 MVP-2). Do not edit or delete — used by downstream tools to gate phase prerequisites.

    Publish the #1352 semantic-design amendment for SEM-235 and API-424, following the existing #1351 design-publication pattern. The delivery defines the accepted contract and migration boundary; it does not claim new wire readers, production orchestration, or backend realization.

    1. Add ADR-111 and a revisioned formal applicability/evaluation addendum. Distinguish the immutable admitted apparatus, required profile obligations, exact-cut applicability and coverage, and proven inapplicability. Specify typed dependency invocation (including A → B → A), explicit false-trigger results, required input closure, admitted support constraints, scoped speculative state, and expected-head atomic commit.
    2. Define unscheduled parent decisions separately from independently authorized effects, execution prerequisites, success-dependent consequences and independent consequences. Preserve deny/withhold under every phase, reject parent/effect dependency cycles, and retain fresh identity, ordinary admission, all blockers and bounded causal work.
    3. Amend SEM-235/API-424 statements and their evidence boundaries, explicitly clarify ADR-108 with its amendment record/pin, and link the addendum from incumbent semantic/concept/reference entry points. Add worked cases and producer/reader migration rules preserving legacy evaluation, claim, state and budget meaning.
    4. Write failing bounded model properties before adding their semantic relation. Reuse the existing SEM-235 typed-result/conjunction oracle where its revision-1 meaning applies. Cover mixed sinks, absent/unresolved obligations, immutable predecessor inputs, dependency-required advice, optional failure isolation, bounded support, false triggers, parent decisions across phases, independent audits, effect cycles, state conflicts and failed commit. Preserve the existing legacy witnesses.
    5. Record exact SEM-235/API-424 path assignments and traceability through the incumbent requirement-scope mechanism, with narrow ownership additions. Verify the policy consumer rejects missing ownership before adding those assignments. Existing ACTIVE statuses remain; amended semantic acceptance is distinct from executable adoption.

    TDD paths: all five issue acceptance criteria and their new semantic relations use Path A bounded executable design witnesses alongside Path D explanatory prose; no documentation-only carve-out is claimed. The published requirements' existing domain/algebra, closed carriers, provenance, visibility, freshness, budgets and historical SEM-233 clauses remain covered by the existing finite SEM-235 and API-424 tests and explicit preservation clauses. These tests make bounded design claims only.

    Security: retain bounded/closed ingress, trusted-context resolution, authority and projection/final-sink gates, safe coarse diagnostics, origin-bound effect authority, and commit/final-invocation fencing. No new parser, launcher, environment or secret surface is introduced. The predecessor view must respect disclosure authorization; a digest does not establish trust.
    Maintainability: reuse current result/graph/conjunction concepts, ADR-104 store authority, API-407 support semantics, closed effect owners, ADR-059 pin tooling and requirement governance; do not duplicate runtime logic.
    Extensibility: bind applicability, obligation coverage, slot dependencies, state scope, support constraints and effect prerequisites by revisioned identities, keeping new sinks/providers independent of backend-name branching.
    Whole-repository impact: normative specs, ADR index, requirement records, reference docs, migration docs, bounded Python tests, requirement ownership/scope, targeted policy checks, pre-publish hygiene, CI and Sonar. No schema, package-version, changelog or release-history edit is planned.

    Validation: set RAES_REQUIREMENT_UID=SEM-235; run the new explicit pytest modules plus test_sem_235_modular_control.py and test_api_424_composition_derivation.py; use file-local repo policy, targeted requirement-scope/ownership, ADR immutability/template checks, changed-file Ruff and link checks. Baseline: 47 existing targeted tests and initial file-local policy passed. Full test/integration/fuzz/completion/policy suites remain in CI. Run the required pre-push review, publish hook, CI/Sonar and readiness gates before requesting merge.

  6. Brad-Edwards commented on Sep 22, 2026

    @Brad-Edwards
    CollaboratorAuthor

    gc_codex_review — sanitized deferred publication for issue #1352, cycle 1 of 1
    Reviewed revision: 332e3e2ae74edabc42dd823f56648d034e2c9e93795fe128cf6d3d147817c948

    Architectural read

    The semantic amendment separates apparatus identity, obligation coverage, dependency inputs, parent decisions and independently authorized effects. It reuses incumbent composition and atomic-store authorities, preserves historical interpretation, and states the finite model's limits. Both core and security review found no actionable issues in the complete 19-file diff.

    Verdict: ship

    Findings

    • (none)
  7. Brad-Edwards commented on Sep 22, 2026

    @Brad-Edwards
    CollaboratorAuthor

    gc_codex_review pre-push cycle 1 of 1 complete for issue #1352 on branch '1352-control-applicability-decisions'. Posted by the MCP server to enforce the pre-push hard-cap-1 contract (issues #796, #804, #906). Do not edit or delete — used by the next gc_codex_review (uncommitted) invocation to count cycles.

  8. Brad-Edwards commented on Sep 22, 2026

    @Brad-Edwards
    CollaboratorAuthor

    Review decision record — codex cycle 1 (issue #1352)

    Reviewer: codex
    Cycle: 1
    Verdict: ship

    Architectural read:

    The semantic amendment separates apparatus identity, obligation coverage, dependency inputs, parent decisions and independently authorized effects. It reuses incumbent composition and atomic-store authorities, preserves historical interpretation, and states the finite model's limits. Both core and security review found no actionable issues in the complete 19-file diff.

    Blocking findings: 0 (clean run)

  9. Brad-Edwards commented on Sep 22, 2026

    @Brad-Edwards
    CollaboratorAuthor

    Pre-PR base synchronization

    • Source: refs/remotes/origin/dev at c3a9154b9be85f4cf60688df9c384385baae03cd
    • Outcome: already_current
    • Published feature head: 7d0d8c34e147ba0641ee1c38e8b4da67911ad1b0
    • Synchronized tree: 3df3f021b2d847bfc9aaa45aea4ae6766c34277f
  10. Brad-Edwards commented on Sep 22, 2026

    @Brad-Edwards
    CollaboratorAuthor

    Pre-PR base synchronization

    • Source: refs/remotes/origin/dev at c3a9154b9be85f4cf60688df9c384385baae03cd
    • Outcome: already_current
    • Published feature head: b57796ae0ff6b3f4d02be653b51ce59c8ace7e88
    • Synchronized tree: 419dcf9d04a00723368aa474a23ead9457b2d580
  11. Brad-Edwards commented on Sep 22, 2026

    @Brad-Edwards
    CollaboratorAuthor

    Execution obligation PHASE-E-WORKFLOW-ON-DEV — Opened

    Category: workflow
    Observed state: Mechanical readiness refused with phase_e_workflow_missing: .github/workflows/ground-control-phase-e.yml is absent from the trusted dev base revision.
    Impact: No trusted delivery handoff or ready-for-review record can be published until the workflow is present on dev.
    Current obligation: Prepare and verify the official pinned Phase E workflow as a separate bootstrap delivery, obtain the required user merge into dev, then resolve this obligation and rerun readiness for #1352.

    Evidence

  12. Brad-Edwards commented on Sep 22, 2026

    @Brad-Edwards
    CollaboratorAuthor

    Execution obligation PHASE-E-WORKFLOW-ON-DEV — Escalated

    Category: workflow
    Observed state: PR #1378 is implemented, reviewed and passes hosted checks, but readiness requires Phase E on dev. Prepared bootstrap PR #1381, whose official workflow is rejected by the repository permission policy.
    Impact: The official post-merge finalizer requires issues write access, while current repository policy classifies every pull_request job as untrusted. The workflow cannot land with green CI without an explicit policy design repair.
    Current obligation: Resolve the GOV-913 policy conflict under reviewed implementation scope, merge the verified bootstrap into dev with user authorization, then resolve this obligation and rerun #1352 readiness.

    Evidence

    Pause class: significant_security_decision
    Decision request: Authorize /implement 1380 to review a narrowly bounded post-merge finalizer permission exception; no broader PR write-permission relaxation.

    This obligation remains open while the decision is pending.

  13. removed
    in-progressAn agent is actively working this issue via /implement
    on Sep 22, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    area:runtimeRuntime and control-plane codedocumentationImprovements or additions to documentation

    Type

    No type

    Projects

    No projects

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions