tegwick
96d11431dc
Assistant: codex Assistant-Model: gpt-6-astra Assistant-Session: 01a0cffe-206f-7323-89d3-b20ca4cf4bf9
1.7 KiB
| title | status | maturity | families | scales | admission | ||||
|---|---|---|---|---|---|---|---|---|---|
| Invalidate and Reproject | draft | hypothesis |
|
|
research-candidate-only |
Invalidate and Reproject
Context: several views derive from remotely accepted records. Problem: a successful write does not refresh every observation. Freshness competes with network work and interaction continuity.
Proposed solution: identify affected observations after acceptance and reload or replace them from authoritative data. Guard against obsolete responses and preserve unfinished drafts separately. Show accepted-but-refresh-failed when necessary; a failed refresh does not undo the accepted mutation.
Consequences: views can converge without treating speculative data as truth, but invalidation scope and ordering need explicit contracts. Blind cache replacement loses edits; broad refresh costs bandwidth. A continuous subscription or authoritative response containing all affected data may avoid a separate fetch.
Evidence: T04 describes React Router action revalidation and stale response handling, including its limits. One framework mechanism, not independent application recurrence. Exact known uses of this whole draft: unknown. Hypothesis only.
Proposed connection: depends on outcome information from Mutation Outcome Tracking and preserves Local Draft State. Client ordering cannot replace server checks.
T01–T05 refer to TransactionalEvidenceExpansion. Connections are research proposals, not admitted graph edges. Reasoning exercise: VC-01.