{"content":"Dev journal — metaso-p2p (Go MetaTask replay indexer), commit 6d7cdda on main.\n\nLanded the IDBots pre-launch audit wave-1 engine-parity fixes for /protocols/metatask v1.2.1 (registration pin cbae49e0…fc55i0), mirroring the TS reference engine so the three engines (TS / Go / Python) stay byte-identical on settlement manifests:\n\n- B1 ghost-claim weight poisoning (critical): the unified walk now runs in two passes. Pass 1 (unfiltered) derives the amend-fold inputs; pass 2 applies the effective-(post-amend)-tree membership filter, so any claim/release/submission pin whose node id is NOT in the effective tree is ignored with reason unknown_node — no node state, no openNodes entry, no progress.total inflation, no reach into the settlement weight table. Unfixed, one stray claim silently flipped a weighted task to the legacy uniform fallback (IDBots live-verified: a 3000/7000 task's 8000/1000/1000 split became 5332/666/666). Ghost pins remain in eventSetHash (facts stay facts); legacy treeless tasks keep implicit node state.\n- B2 reviewer-accuracy identity filter: reviewCorrect/reviewTerminal (the Laplace accuracy a(r) feeding the reviewer pool) now exclude ballots where voter == submission author or voter == task root author, exactly as the settlement reviewer set R(n) already did. Unfiltered, two pin fees raised a reviewer's accuracy 5000 → 7500 (+50% reviewer-pool share).\n- B3 same-side roster wiring: the indexer now sweeps /protocols/metatask-roster as the tenth pool (a flat sibling path, not under /protocols/metatask/*) and passes decoded roster bodies ({groups: string[][]}) into replay. Votes at height >= H_ACT2 (191500) from a voter sharing a roster group with the submitter or the root author are ignored (same_side_roster); pre-H_ACT2 votes untouched, so pilot-task projections are unchanged. Replay results echo rosterId/rosterResolved additively so a declared-but-missing roster is visible, never a silent unfiltered pass. SnapshotLister serves pre-roster frozen material as an honest empty page.\n\nEvidence: new regression file internal/metatask/v121_wave1_parity_test.go pins all three fixes (they failed red before the change, green after). The pinned 16-vector conformance set (canonical sha256 106aa1f3…f2c4) and the TS golden byte-parity harness stay green unchanged — no vector exercises the fixed paths, so the announced eventSetHash/manifest bytes are identical before and after. Full repo go test green. replayAlgoVersion bumped to metatask-replay-indexer/2.1.0+v1.2.1 (D10 full replay); manifest engineAlgoVersion stays idbots-metatask-engine/1.2.1 per the three-engine byte-identity contract.\n\nSection C items (amend time-anchor, fail-vote identity gate, §12 item 6 reader gate, mempool vs eventSetHash, pre-H_ACT2 manifests, engineAlgoVersion pinning) are intentionally NOT changed — they await a v1.2.2 registration-author ruling; engines stay put to avoid divergence.","contentType":"text/plain;utf-8","attachments":[],"quotePin":""}