Skip to content

Clarify module storage, scoped calls and verifiable recovery guarantees - #7

Closed
plur9 wants to merge 27 commits into
mainfrom
fix/module-resolution-audit-20260913
Closed

plur9 wants to merge 27 commits into
mainfrom
fix/module-resolution-audit-20260913

Conversation

@plur9

@plur9 plur9 commented Sep 13, 2026

Copy link
Copy Markdown
Contributor

Compatibility follow-up: proposed workflow policies are opt-in and default off.
Existing module data paths and unqualified callable names are retained where valid;
scoped names are optional. See BEHAVIOR-OPTIONS.md on the source audit branches.
Safety corrections remain active. Instance-bound execution and its DIP changes
remain parked under datacore-one/datacore#192.
Legacy advisory allocation compatibility and the full-mode core/runtime coupling
still require qualification. Unknown-effect retry policy is awaiting an owner
decision; this change does not alter its current duplicate-effect protection.
This remains a mixed draft audit PR, not approval to merge or deploy the parked model.

Architecture deferred (2026-09-14): The instance-bound execution / mandatory admission / no automatic takeover or retry model is parked for future design review in datacore-one/datacore#192. Preserve this branch and its proposal. This PR also contains independent audit fixes; separate and verify those before merging, rather than merging the deferred model with them. Deferral does not certify deployment or close demonstrated defects.

The module specification placed private state inside shareable scoped code and left callable scope and workflow execution evidence ambiguous. DIP-0022 now defines a separate canonical private state root, preserved migration/refusal semantics, qualified scoped calls, and the distinction between diagnostic state and actual execution.

The branch also retains the audit amendments to Draft DIP-0046 for exact publication acknowledgements, preserved recovery/moves, and outgoing-history authorization. Draft DIP-0049 now describes verifiable module resolution and observed registration health. Both remain Draft; future functionality is not promoted to a current guarantee. Material changes have traceable requirement/problem/compatibility/test/deployment records.

DIP-0009 also records proposed cadence capture/history/completion and continuous orchestration invariants: machine-origin proposals require review; capture does not acknowledge completion; successful cadence history requires durable task/output proof; legacy state and interrupted publications are preserved. Heartbeat proposals enter the shared review/allocation path; health and command results require validated completion evidence, and corrected contract holds return to current review. Change-control records distinguish the historical Implemented specification from unmerged amendments and future DIPs.

DIP validator diagnostics now redact private values, with regression tests and pinned read-only CI actions. The proposed heartbeat-state amendment requires preserved stale and malformed evidence, explicit decision updates, declared writer attribution and canonical migration, with compatibility and test changes recorded.

The proposed configuration amendment requires consistent bounded policy interpretation, complete canonical venture discovery, explicit observation controls and honest portfolio evidence. It retains extension fields and records compatibility and deployment impacts.

Further proposed amendments define preserved monthly budget history, stale-write and retry semantics, honest declared-writer heartbeat diagnostics, and consistent UTC cadence scheduling. They distinguish snapshot budget checks and diagnostic status from actual external-spend or execution admission.

Validation at 6984092: 45 DIPs passed; validator tests passed (2 methods, 15 subtests). Related implementation: core PR191 and MCP PR21.

Draft review only: runtime rollout and the full audit have not converged. This PR does not ratify draft DIPs or claim completed deployment.

The proposed hypothesis evidence amendment records preservation across historical representations, configured source consistency, declared writer history, safe retries, explicit unknown states, durable local mutation and its cross-host limits. It documents the pinned YAML dependency and matching deployment requirements. All 45 DIPs validate and both validator tests pass. No implemented/audited status is promoted.

Checkpointf34f6ef adds traceable proposed amendments for shared installed module-reader selection/failure semantics and canonical pending-cadence identity. Observation cannot suppress a different role/frequency by name or bypass legacy reconciliation; diagnostic counts use validated authoritative evidence. All45 DIPs validate, validator tests pass, and no DIP status is promoted.

plur9 added 26 commits September 11, 2026 13:16
@plur9

plur9 commented Sep 15, 2026

Copy link
Copy Markdown
Contributor Author

Superseded by #8, which carries the same audit amendments with the parked execution model removed.

This PR cannot merge as it stands: it contains the instance-bound execution proposal the owner parked, and that proposal must not reach main (datacore-one/datacore#192). #8 removes 1,241 lines of it across DIP-0011, DIP-0041 and DIP-0009 while keeping everything compatible — including Canonical alias compatibility, which sat inside the removed block but is about DIP-0015 space discovery and would have reinstated an automation break if dropped with its parent.

Closing rather than leaving open, because no amount of review makes this branch mergeable under the current decision. Nothing is lost: the branch stays, and datacore-one/datacore#192 records the preserved checkpoint (06feab9) for later review.

scripts/validate_dips.py on #8: OK — 45 DIPs valid, numbers unique, hygiene clean.

@plur9 plur9 closed this Sep 15, 2026
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