tegwick
96d11431dc
Assistant: codex Assistant-Model: gpt-6-astra Assistant-Session: 01a0cffe-206f-7323-89d3-b20ca4cf4bf9
1.6 KiB
| title | status | maturity | families | scales | admission | ||||
|---|---|---|---|---|---|---|---|---|---|
| Version-Checked Mutation | draft | hypothesis |
|
|
research-candidate-only |
Version-Checked Mutation
Context: a proposal depends on a previously observed record. Problem: concurrent updates can invalidate that base. Consistency competes with throughput and user effort.
Proposed solution: send an expected version with the proposal; the authority checks it atomically with validation and commit. A mismatch rejects the attempt. Preserve the draft, fetch current state and reconsider intent before a new attempt. An operation identity for safe retries is separate from the version precondition.
Consequences: lost updates can be detected, but high contention creates repeated review. Unconditional or commutative operations may need different preconditions; blindly retrying with a new version can overwrite another user's intent.
Evidence: T01 documents Kubernetes conditional PUT; T02 documents a different GitLab failure surface and optional last_commit_id, insufficient to prove identical atomic semantics. One precise mechanism use, not two verified frontend uses. Recurrence of this frontend configuration: unknown. Hypothesis only.
Proposed connection: rejection creates work for Local Draft State and Mutation Outcome Tracking.
T01–T05 refer to TransactionalEvidenceExpansion. Connections are research proposals, not admitted graph edges. Reasoning exercise: VC-01.