docs(book): explain advanced Akita relation layouts - #349
Conversation
PR SummaryLow Risk Overview The new page frames three layout axes on top of the four unchanged physical relation families, then fully documents multiple commitment groups: per-group folded responses and local Basic relations in an Akita fold now points readers to the advanced chapter instead of saying advanced layouts are out of scope, and the verifier section cross-links the generalized logical layout before the opening-points witness-order page. Reviewed by Cursor Bugbot for commit 022ed79. Bugbot is set up for automated code reviews on this repo. Configure here. |
Documentation blast radius (advisory)These regions may need doc/spec/book updates based on changed paths. Changed files in this PR: 3
|
Benchmark Report
Negative deltas are improvements for time, memory, and proof size. Terminal response component breakdown
The Detailed schedule and proof-size breakdowns by fold level are available in the uploaded |
CI test timing
Run summary
Wall time spans 2 parallel nextest slice shards. Slowest tests
Regressions vs mainNo per-test regressions above the threshold. New slow testsNo new tests ≥30s vs main baseline. |
…d' into codex/advanced-objects-intro
…k-advanced-relation
…k-advanced-relation
quangvdao
left a comment
quangvdao
left a comment
There was a problem hiding this comment.
Freya, thank you for the thoughtful stacked follow-up. The final-group-first ordering, group-local consistency/A/B structure, level-shared D relation, and transition back to a single recursive witness are all valuable explanations. I also appreciate that the chunk and mixed-dimension sections are clearly marked as scoped stubs rather than presented as finished derivations.
I am requesting changes on the advanced chapter for three code-level mismatches:
- Groups do not share one root opening point. Claims within a group share a point, but each PolynomialGroupClaims owns its complete point. Different groups may have distinct values and arities; the heterogeneous-group end-to-end test proves and verifies exactly that case. Please revise the shared-root-point statement and carry group-local points through the explanation.
- The stated multi-group RHS and witness are the raw-mode realization, not the production root layout. A production root is compressed: ordinary B/D targets are private intermediate images, F/H compression rows are appended, only terminal p_F/p_H payloads are public, and WitnessLayout contains compression digits, alignment ranges, and compression quotient rows. Please label the existing equations as the raw or semantic layout and add the compressed multi-group physical layout inherited from #340.
- The source list cites crates/akita-verifier/src/protocol/ring_switch/mixed_relation.rs, which was deleted. The current verifier implementation is in relation_evaluation.rs and prepared_relation_point.rs. Please update the citation and fix scripts/check-book-source-paths.sh so this class of missing path fails CI; the current rg output includes the Markdown filename prefix, which the script mistakes for the cited path.
The existing opening-points-layout.md page also describes only a simple ordinary quotient tail. Once #340 corrects that foundational page and folds/archives the compression and Stage 2 specs, please rebase this PR and make the advanced chapter use the corrected mode-aware terminology. The spec lifecycle work belongs in #340, so it does not need to be duplicated here.
The structure and exposition are promising. These are targeted corrections needed so readers can safely treat the new chapter as a description of the shipped protocol.
|
Coordination note from PR #371: this PR remains the owner of advanced logical relation layouts. PR #371 now documents verifier replay, exact unequal chunk ranges, mixed setup roles, compact trace and setup evaluation, and terminal checks. When this stack is updated, please keep the logical multi-group exposition here, but use a distinct complete opening point for each group, describe the compressed F and H production rows, and replace equal-size or padding assumptions with the exact ranges |
2 participants
Summary
Scope
This is a stacked follow-up to #340. The base PR owns the one-group, one-chunk, common-dimension derivation; this PR owns the generalized physical layouts.
Testing
Dependency
This PR targets freya/akita-book-relation-in-fold and should be reviewed after #340. Once #340 merges, it can be retargeted to the parent book branch.