G.9:4.3 — Execution protocol (run‑time / selector‑adjacent)

Preface node heading:g-9-4-3-execution-protocol-run-time-selector-adjacent:140620

What this page is

This is generated FPF reference text from the specification preface or supporting sections. It helps interpret FPF; it is not FPF Reference product documentation.

Methodology

Use it to understand how the specification wants to be read, then return to a route, pattern, or work packet for active work. Cite generated IDs only when the wording changes the task decision.

Content

For one run under the pinned plan, follow these steps:

  1. Validate CSLC references and pins. Validate the cited CSLC comparability and admissibility references, active pins, and witnesses. Apply the eligibility or acceptance checks required by the pinned comparison and refuse or abstain on non-admissible operations. When using G.5, record its branch trace in its own result vocabulary and use the planned TaskSignatureRef. Add the matching TaskMapRef when that selector uses G.4 CAL gates. If a live A.21 gate consumes this check, cite its GateDecisionResult and add a DecisionLog only when needed for audit or reuse. The profile application and its action consequence remain governed by A.21; do not create a G.9-local CSLC gate.

  2. Run the comparison and any required selection. Apply the pinned comparator to the admissible comparison inputs and retain its lawful result shape in ParityReport. If a selector-facing result is required, apply G.5 under the plan's exact task and conditional map refs and publish its declared outcome. An ordinary parity or benchmark report ends without G.5 when no selector result is needed.

    Under componentwise minimization, A=(8,3) and B=(10,2) trade cost against time. An admitted comparison can return the singleton {A ∥ B}. If A and B are all eligible candidates, both are nondominated. Add C=(7,2) and the frontier becomes {C} while the relation between A and B stays the same. Under a comparator that admits intervals, take A=(8,[3,4]) and B=(10,2). Every allowed value for A’s time still gives incomparability. With A’s time in [1,4], A may dominate B or the two may trade off. Report only what the approved comparison supports. If two bounds used as evidence about the same profile contradict each other, resolve that conflict before reporting a relation. If the profiles are consistent and none meets the requested limits, that can settle the eligibility question. Separately, if a required calibration is missing, apply the evidence policy. An admitted narrower comparison can return only its justified tokens. An abstain decision leaves the comparison slot unbound.

    For a bounded-specialization comparison, the report should include the active specialization profiles or equivalent pins so readers can recover the work-measure threshold target, prior exposure, budget-to-threshold, post-threshold efficiency when relevant, transfer, retention, downside field, and any corridor-entry baseline or evidence note from the report.

  3. Record the comparability mapping when used. If UNM_id?, NormalizationMethodId[]?, or NormalizationMethodInstanceId[]? was declared, echo it in ParityReport or its explicit pins delta. Record the ids and any scoped notes required by the cited specification in the audit pins and SCR; cite the applicable PathId values.

  4. Publish trace. Emit ParityReport with the ParityPlanRef, its BaselineBindingRef, conditional selector input refs, EvidenceGraph citations, and all active edition and policy-id pins, so the run can be checked and run again.

  5. Emit telemetry for refresh when produced. Publish the telemetry pins or events used for refresh. Treat the values as report-only unless a CAL policy promotes them and its ID is recorded.


Last Updated: 2026-11-22 — upstream FPF commit 7a93d8f8 (github.com/ailev/FPF)