Repository navigation
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub. |
This branch has not been deployed
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Implements read-only grouped remote datasets alongside the existing application-managed window mode, following the approved design in #630.
Completed and reviewed:
Task 1: 11 paired local/manual compatibility tests and recorded geometric browser baseline. Existing runtime/windowing behavior unchanged.
Task 2: typed provider/query contracts, canonical identities, runtime capability checks, session/version separation, aggregate provenance and summary capabilities, explicit unloaded/loading/error slots, public declarations and ES2022-only consumer tests.
Task 3: bounded immutable/revision-fence reference providers, exact grouped ranges and same-version summaries, independent atomic export captures, and deferred transport fixtures. Both reviews passed; 18 website provider tests and 71 remote-core tests passed.
Task 4: sparse grouped projection, bounded height measurements and logical/DOM mapping prototype, with reproducible Chromium/WebKit height-limit and viewport evidence. Both reviews passed; a max-safe-integer search termination defect was fixed with a regression.
Task 5: transactional response validation and session ownership, atomic root replacement, explicit expiry, scoped revocation, and paired payload/evidence eviction. Reentrant callbacks and malformed envelopes were fixed with regressions before both reviews passed.
Task 6: bounded priority scheduling, deduplication, finite retries, absolute deadlines and physical-request accounting. Late authorization and abandoned session cleanup survive cancellation and disposal. Both reviews passed.
Task 7: atomic cache admission, separate payload/ingress/metadata budgets, protected topology and intent, bounded measurement retention, and shared two-generation accounting. Both reviews passed after correcting mixed-batch capacity diagnostics.
Task 8: public headless remote model with atomic admission, bounded request/view retention, sparse grouped reads and versioned server summaries. Both reviews passed after fixing retry, viewport admission and loading-time replay regressions.
Task 9 functional slice: shared reads preserve manual windowing; remote adapter adds atomic geometry, stable anchors, incremental height prefixes and input mapping. Both reviews passed; fresh original browser compatibility checks passed.
Task 9b functional slice: immutable geometry reuse, cooperative response preparation, compact validation evidence, deadline/cancellation cleanup, and pre-admitted incoming metadata with atomic credit exchange. Separate functional specification and code-quality reviews passed. Shared payload, ingress and metadata limits remain unchanged.
Blocking performance gate: Task 9 originally measured integrated expansion p95 34.773 ms against the unchanged 8 ms requirement. Task 9b substantially improves diagnostic timings. The historical repeated default-100 series fails the preparation maximum (15.901 ms). Review corrected accounting for preparation within commit-bearing turns; corrected smoke evidence still fails one synchronous-command p95 (8.154 ms). Every run preserves 5,000 resident payloads and 5,001 topology nodes. The host was loaded; quiet-host reproduction and any resulting fixes remain required. Neither series certifies performance. React integration must wait for that gate. Measurement-budget integration, remote UI, server export jobs, and release evidence remain incomplete. This remains a draft, stacked on the design branch.
Validation: latest full core suite 356 passed (339 remote); layout 163 and renderer 201 passed. Core build/typecheck/lint, public consumer types, API freshness and whitespace checks passed. Earlier compatibility evidence includes React 2,083, paired local/manual matrix 11, focused React compatibility 23, original Chromium window/eviction 10 and local production website Chromium/WebKit 16 passing tests. Separate specification and code-quality reviews passed for Tasks 1–8 and the functional Task 9/9b slices. Review findings were fixed with regression tests before approval; Task 9b performance certification remains open.
Original local changes remain preserved. See docs/superpowers/plans/2026-11-19-remote-datasets.md and status/remote-datasets/ for task status and evidence. No performance claim is made from the loaded-host functional runs.