Repository navigation
Conversation
Keep per-Cell fenced root selection while uploading bounded application-scoped capture cohorts. Issue signed append windows through fresh enrollment and serialize native proof release with durable closure. Document the remaining bundle-coverage protocol and preserve qualification failures. Bundle-based bucket acknowledgments remain disabled.
forhappy
force-pushed
the
codex/packed-write-publication
branch
from
October 7, 2026 09:04
fc8147b to
c590664
Compare
forhappy
changed the title
forhappy
marked this pull request as ready for review
October 8, 2026 01:11
…dmission" This reverts commit bb2f089.
forhappy
changed the title
forhappy
marked this pull request as draft
October 8, 2026 19:46
forhappy
marked this pull request as ready for review
October 8, 2026 23:18
forhappy
marked this pull request as draft
October 8, 2026 23:47
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Cellule's per-Cell publication work limits write throughput. This draft adds shared node-bundle selection with exact range proofs, signed append grants, bounded authenticated histories, actor receipt/read/retry integration, exact capture retirement and separately admitted asynchronous Cell roots. Authority fencing, origin verification, complete issued-range closure and joined shutdown remain required.
The latest fix observes authenticated selection readiness without owning the Cell publisher. A ready oldest prefix enters exact cleanup; older root debt can prepare while selection waits. The unchanged cleanup deadline begins after readiness. A real-actor regression fails with
Fencedbefore the fix and passes afterward: delayed selection beyond ten seconds after valid Fleet ACKs, read/retry visibility, older-root progress, joined drain, cold state/results and released credits. The atomic gate-ID allocator also preserves checked uniqueness and Rust 1.97 compatibility without the API deprecated by current Clippy.No acceptable throughput improvement or celld parity is established. Six fresh matched release cases compare
4a8f55c, candidate6c909a6and celldf2bf6486. They use 1,000 uniform Cells, 96-byte SQL INSERT+SELECT, a two-hour retry/result ledger, 128 clients/queue slots, 30-second warmup and 60-second window. The VM has 8 CPUs / 8 GiB shared memory; container ceilings exceed it. These overloaded completion counts are not sustainable capacities.Candidate Fleet returns 328,243 errors, drops 560,671 offers and fails 19,797 of 21,366 warm ACK checks; cold recovery and successful aggregate drain are unverified. Candidate Bucket has zero request errors and passes all 26,423 warm/cold ACK checks, but drops 104,823 offers. Celld Fleet is OOM-killed and fails its audit; celld Bucket passes all 124,185 warm/cold checks but drops 40,437 offers. All six points fail qualification. Fleet completion decreases 5.3%; Bucket's +0.46% difference does not establish a gain and bypasses the managed producer. The subsequent atomic compatibility fix was not part of these timed binaries; it is not presented as a performance improvement.
Fleet root density remains 11.12 commands/root; canonical PUT attempts are 0.5422/completed write and GET/range attempts 13.9287. Retention rises from 27.88 to 58.40 MiB of 64 MiB; oldest debt ends at 45,452 ms. Steady Bundle ACKs remain zero. The 215-command / conditional 0.05-PUT targets and bounded-debt qualification remain unmet. Sampled audit failures occur on POST retries after matched GETs; the pre-SQL backlog gate remains a concrete next regression target, not the proven cause of every failure.
Validation: the frozen selection-fix snapshot passes all eleven contributor routes, 1,962 passed / 0 failed / 38 ignored. All eleven contributor routes pass again after compatibility commit
d852525, and Rust 1.99 Clippy passes for all targets/features with warnings denied. Final-head CI remains required. This compatibility change is outside the timed candidate binaries. Canonical reports and independent journal replay reconcile every offer/attempt/completion, error/drop and complete ACK cohort. Fixture, client/auditor, host and runner provenance match. All 9,530 indexed external evidence files rehash; raw journals, binaries and provider data stay outside Git.Latest measurement and merge criteria, earlier repeat, delivery, runtime design.
Remaining before merge: restore pressure-time read/retry availability and successful joined drain; prove materializer progress and bounded debt; connect Bucket shared selection; reduce complete publication cost; qualify failed-owner Fleet suffix recovery and safe collection; finish sustainable capacity search, three paired five-minute repetitions and read-only/mixed guardrails with zero errors/drops and the original latency targets. The 2,000-Cell / 10K-write / 50K-read standard-node goal remains unqualified. Keep this PR in draft.