Repository navigation
docs(D-CFA-1): source-harvested ON/WHEN/DO incidence and refusal gates - #1814
Conversation
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars.
 |
31c3a93 to
bbd8980
Compare
|
Architectural crosswalk correction (operator, 2026-10-27; see #1819): CFA-1's typed ON/WHEN/DO incidence must explicitly admit multiple mappings of the same native algebra/polyfill, and many simultaneously applicable source rules converging on one qualified semantic state. Source MethodId and typed coordinates preserve distinctions, but they are not demands for independent executable method runtimes. A shared ON/DO cell is often normal; it is not itself an ambiguity or last-wins write conflict. The operator's Odoo example: a taxable item can receive converging rule contributions, with subordinate tax-rate rules, and the resulting determination may be routed to two distinct accounting ledgers. Preserve many-rules→one-state and one-state→multiple-destination composition, under source-defined combination/order/rounding/authorization laws. Do NOT hard-code OR, SUM, last-wins, or treat identical local ordinals as equivalent ledger identities. This is a proposed source-attested test fixture; the incomplete #44 Odoo corpus does not establish real tax or dual-ledger semantics. Suggested CFA-1 falsifiers: (i) two separate, valid, conditionally mapped registrations to one polyfill jointly contribute to a shared result; removing either mapping from a projection by accidental dedup is RED, without deleting actual source methods; (ii) route one business result into two typed ledger outputs, and ensure neither destination is lost or aliased; (iii) where combination law/guard/source scope is not attested, UNKNOWN/REFUSE rather than assume an operator. #1819 now documents the candidate factored Quack lowering and these fixtures; it does not replace CFA-1 as logical authority. Also recognize that conflict is a signal but bounded Thought-Git synthesis is an explicit goal. Its 0+63 hypothesis slots compare alternate dimensional valuations/source worlds to propose conditional compositions; not automatic write union or proof. No change requested to runtime or ownership. |
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
…atch fold #1814 and #1819 have merged, so the plan now cites them by section from this repo instead of as open PRs by line number. That line-number row was the one citation-decay failure on c440fab: the file did not exist locally, so the local run reported UNVERIFIABLE where CI reported DECAYED. Reading both in full showed two contradictions, now fixed: - the dispatch step's two-valued `OR over w in G` fold is restricted to candidate DO presence; selection evaluates typed guards in K3 (#1814 kill test I03); - open question 3 now follows #1819: several mappings and convergence are normal; a conflict needs incompatible source-qualified results. Also corrects the quoted Ruff role names to #1814's list. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMNiyjwqFke9CfmoBgn
…GIT reference predicates (#1831) * docs(D-DT64-6): DR64 routing, DT64 bind store, per-cell WHEN×ON×DO, OGIT reference predicates Records seven user decisions from 2026-10-28 in DECISION / SCOPE / BASIS / REVISIT WHEN form: - the three DT64 levels; - DR#3 Identity × Properties as the primary view; - WHEN×ON×DO per DR#3 cell; - DR64 = direct routing and DT64 = DTO bind store; - immutable classid → behaviour via hotplug; - global behaviour needs no classid; - OGIT read unpinned. Maps each decision onto existing code and plans with evidence states, and stages D-DT64-6-0..5: - OGIT reader reconciliation; - ogit:allowed expansion in lance-graph-ontology; - DT64 predicate registration; - consumer HotPlug access. Requests, but does not apply, an amendment to D-DT64-Q0 line 9. Docs only. No code, ABI, test, runtime or merge authorization. Board: entry, INTEGRATION_PLANS and STATUS_BOARD prepends, and the regenerated entries index. The supersession index regenerates unchanged. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMNiyjwqFke9CfmoBgn * docs(D-DT64-6): cite in-repo sources by symbol and section, not line number The citation-decay gate failed on nine grounding rows. The cited lines were correct, but each row's citation sat next to the following row's backticked text, which the gate took as its anchor. Line numbers into moving files are what the gate exists to retire, so every in-repo citation now names a symbol or section heading instead. Also replaces the two grounding rows that rested on grep output with claims from full reads: the three merged DT64 plans, and the qr_focus / qr_register lab tests (whose actual wording is now quoted correctly). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMNiyjwqFke9CfmoBgn * docs(D-DT64-6): record D8, DR64 references identity and the register copy amortizes prefetch DR64 direct routing references existing identities and mints none. The global->local prefetch holds a copy of the register (identity references and local positions only, never values), reused by every read under one frame and rebuilt when the frame check refuses. Stated against the zero-copy law's pointer-table carve-out and Q0's accounted prefetch cost. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMNiyjwqFke9CfmoBgn * docs(D-DT64-6): ground #1814 and #1819 as merged; K3 note on the dispatch fold #1814 and #1819 have merged, so the plan now cites them by section from this repo instead of as open PRs by line number. That line-number row was the one citation-decay failure on c440fab: the file did not exist locally, so the local run reported UNVERIFIABLE where CI reported DECAYED. Reading both in full showed two contradictions, now fixed: - the dispatch step's two-valued `OR over w in G` fold is restricted to candidate DO presence; selection evaluates typed guards in K3 (#1814 kill test I03); - open question 3 now follows #1819: several mappings and convergence are normal; a conflict needs incompatible source-qualified results. Also corrects the quoted Ruff role names to #1814's list. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMNiyjwqFke9CfmoBgn * docs(D-DT64-6): ON×WHEN×DO is in merged research docs, not absent The plan and board entry said the merged plans do not use ON / WHEN / DO. That holds only for the three DT64 plans. ON×WHEN×DO is established in merged lance-graph research docs: the D-CFA series (docs/research/counterfactual-behavior/) and the EAW series (docs/research/epistemic-awareness/), plus #1819. All twelve files read in full; the grounding row and the non-goal line now say so. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMNiyjwqFke9CfmoBgn * docs(D-DT64-6): place the plan in the #1760 successor design (crai-rs CR-H0/H1) #1760's D-SAG-6 plan is unchanged since merge; its successor is the crai-rs CR-H0 / CR-H1 map (crai-rs #1, #2), which makes D-SAG-6 the motor and Ruff + DT64 the transmission. The plan now grounds both, places the per-cell ON×WHEN×DO registry as that transmission's behaviour binding, ties D5/D8 to D-SAG-6 F06, F07 and F11, and requires behaviour registrations to reference ordered effects: a DO set or deduplicated SPO bag cannot distinguish two effect orders. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMNiyjwqFke9CfmoBgn * docs(D-DT64-6): ground odoo-rs #46 as the Odoo intake to the per-cell registry Searched odoo-rs, OGAR, crai-rs, lance-graph and ndarray for the #1760 lineage. Besides crai-rs CR-H0/H1, odoo-rs #46 (Odoo Codegraph harvest schema, docs) is the one successor; OGAR and ndarray have none. Read #46's doc and schema in full. Its `dt64_candidates` array (source nodes -> existing algebra_ref, no seat mint) is the registry's Odoo intake; every emitted_by MethodId is kept; effective behaviour needs a registry_snapshot in the frame. Open question 2 (inheritance before the cell lookup) is narrowed by #46's _inherit / _inherits rules. Notes that #46's doc lists dt64_candidates as a graph layer while its schema does not. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMNiyjwqFke9CfmoBgn * D-DT64-6: address review on #1831 (OGIT readers, D-DT64-6-1a, Q0 contract) - The two OGIT readers see disjoint verb sets: ttl_parse `classify` accepts the 12 rdfs:Class verbs (all WorkOrder, all ogit:from-to), OGAR `parse_verb` the 511 owl:ObjectProperty verbs, of 523. Measured over an unpinned OGIT clone (b5731a8, receipt only). - Add D-DT64-6-1a: classify every verb file in both readers first. - D-DT64-6-2 builds (src, verb, dst) from the entity ogit:allowed blocks (960 files, none in verbs/) plus the 12 from-to verbs, reads OGAR's OgitTables::Allowed instead of adding a second parser, and puts the standing parity test beside tests/ontology_cache.rs. - 518 / 2,756 is OGAR #363's reader output (eaw-5), not OGIT drift. - crai-rs CR-H0/H1 is a sidecar consumer of #1760, not its successor; its open-PR list is dated 2026-10-27. - Ruff has no Rust frontend; role names checked in ruff role.rs. - Cite D-DT64-Q0 by section, not line 9; INTEGRATION_PLANS says eight decisions. Docs only. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMNiyjwqFke9CfmoBgn * D-DT64-6: align STATUS_BOARD rows with the revised plan Add the D-DT64-6-1a row and restate D-DT64-6-2 as reading OGAR OgitTables::Allowed rather than a second expansion in lance-graph-ontology, matching plan §5 after the #1831 review. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMPWTwovjYJEUMuGPb4 --------- Co-authored-by: Claude <noreply@anthropic.com>
Research question
Can harvested OGAR/Ruff/Odoo method facts become a typed ON×WHEN×DO logical incidence view without a duplicate registry, invented guard/parameter semantics, loss of coexisting methods, or confusion with the current DT64 predicate catalogue?
Adds only:
cfa-1-typed-incidence.mdwith source/identity/guard/multiplicity ownership and I01–I08 falsifier seeds.Stack: base D-CFA-0 draft branch; after review, prior stage lands first. No code, tests, framework or ABI added; no evidence of shipped DR64³ claimed.