[AdaWorldAPI]

Summary

This PR adds a plan, .claude/plans/d-dt64-6-routing-bindstore-behavior-v1.md, and a board entry. The plan records eight user decisions from 2026-10-28 in DECISION / SCOPE / BASIS / REVISIT WHEN form. It ties each decision to existing code, plans and merged research docs, and gives each tie an evidence state. It stages the follow-up work as D-DT64-6-0 to D-DT64-6-5, with D-DT64-6-1a in front of the reader comparison.

The PR is documentation only. It adds no code, ABI, test or runtime, and authorizes no merge of anything else.

Decisions recorded

1. DT64 has three logical levels, each 64×64×64 and reached by focused global→local addressing:

   • DR64³ Identity(+value) × Property(+value);
   • Identity × Property × Predicate;
   • Predicate³ = WHEN × ON × DO.

   The physical third axis is still open (NFR-OP-2).

2. The primary view is DR#3, Identity × Properties. Identity × Identity and Properties × Properties are optional derived views.

3. WHEN × ON × DO attaches to each DR#3 cell as the registry over its methods.

4. DR64 is direct routing (a mapping table). DT64 is the DTO bind store for every holographic address. Behaviour is registered as classid → polyfill over mask-risc / ndarray.

5. A classid's behaviour is permanently immutable. It is versioned under a fixed concept through a derived DTO classid, registered via lance-graph-contract/src/hotplug.rs.

6. Global behaviour needs no classid and lives in lance-graph. Consumers get a classid through the hotplug pattern.

7. OGIT is immutable and read unpinned.

8. DR64 references existing identities and mints none.

   • The global→local prefetch keeps a copy of the register, and every read under one frame reuses it.
   • The copy holds identity references and local positions only, never values.
   • When the frame check refuses, the copy is rebuilt.

OGIT: the two readers see disjoint verb sets

Measured over an unpinned OGIT clone (commit b5731a8, recorded as a receipt, not a pin), with three sample files read in full:

	Count
Verb files	523
a rdfs:Class; rdfs:subClassOf ogit:Verb (all NTO/WorkOrder/verbs/, all with ogit:from-to)	12
a owl:ObjectProperty; rdfs:subPropertyOf ogit:Verb	511
Files with ogit:allowed	960, none under a verbs/ directory

• lance-graph-ontology ttl_parse classify accepts only the rdfs:Class form, so it sees 12 verbs. The other 511 fall to SubjectKind::Other.
• OGAR parse_verb accepts only owl:ObjectProperty, so it sees the 511 and rejects the 12.
• (src, verb, dst) signatures live in the entity files' ogit:allowed blocks. OGAR extract_allowed_block reads them, and OgitTables::load keeps a tuple only when both ends resolve. Every other tuple is reported as Unresolved.
• 518 verbs / 2,756 edges (eaw-5) is OGAR Super-domain RBAC + UnifiedBridge: spec + Tier A starter (D-SDR-1, D-SDR-2) #363's reader output. It is a reader difference, not OGIT drift, and is never merged with 523 / 2,894.

What follows for the plan:

• New D-DT64-6-1a: classify every verb file in both readers before comparing them.
• D-DT64-6-2: build the table from the entity blocks plus the 12 from-to verbs. OGAR's OgitTables::Allowed is the default source, so no second parser is added. A standing parity test goes next to OGAR:crates/ogar-action-handler/tests/ontology_cache.rs.

Relation to #1760 (D-SAG-6)

I searched odoo-rs, OGAR, crai-rs, lance-graph and ndarray. The D-SAG-6 plan has not changed since #1760 merged, and OGAR and ndarray carry nothing related.

crai-rs CR-H0 / CR-H1 (crai-rs #1 and #2, private repo) consume #1760; they are not its successor. They describe a sidecar retrofit around the D-SAG-6 motor:

• D-SAG-6 stays the motor and keeps ownership.
• Ruff harvest plus DT64 versioned bindings are the transmission.
• docs(D-CFA-1): source-harvested ON/WHEN/DO incidence and refusal gates #1814's ON×WHEN×DO is the source-intake and guard ledger.

CR-H0 lists #1813–#1830 as open "as checked 2026-10-27". #1814 and #1819 have merged since, so that list is out of date.

Ruff has no Rust frontend. Its SPO crates on main cover C++, C#, MeTTa, Python, Ruby and SQLAlchemy, so this transmission cannot harvest Rust sources yet. The role names #1814 cites are PredicateRole bits in ruff_spo_triplet/src/role.rs, and I read them there.

odoo-rs #46 (docs, plus an optional JSON Schema and an example) is the Odoo intake to the per-cell registry:

• dt64_candidates maps source nodes to an existing algebra_ref, with no seat mint and no DO authority.
• Every emitted_by MethodId is kept. fix(cam_pq): add missing arrow::array::Array import in udf.rs #44's single-emitter map lost 26 / 37 edges.
• Effective behaviour requires a pinned registry_snapshot, and that snapshot is part of the D8 frame.
• Its _inherit / _inherits rules narrow open question 2.

This plan is the transmission's behaviour binding. Three D-SAG-6 gates apply:

Gate	Rule	How this plan meets it
F07	A DR ordinal from another frame is refused	D8's register copy is keyed to its frame
F11	A changed source SHA under a retained recipe id must invalidate or rebind	D5 re-routes to a new derived classid
F06	Unexamined regions are UNKNOWN	An unregistered cell is UNKNOWN, never "no behaviour"

A behaviour registration must reference ordered effects, not just a DO set. *x = 1; *x += 1 and its reverse have the same reads and writes but different results.

Where ON × WHEN × DO already stands

• The three DT64 plans do not name ON / WHEN / DO.
• The triple is established in merged research docs:
  • D-CFA (docs/research/counterfactual-behavior/, incl. docs(D-CFA-1): source-harvested ON/WHEN/DO incidence and refusal gates #1814);
  • EAW (docs/research/epistemic-awareness/);
  • docs(research): factor ON×DO dispatch through witnessed Quack WHERE populations #1819.
• As a registry, it exists in the odoo-rs lab design docs/DT64-ON-WHEN-DO-REGISTRY.md (fix(cam_pq): add missing arrow::array::Array import in udf.rs #44 / feat(planner): dynamic elevation — cost model that smells resistance #45, test-only).
• The plan's guard selection evaluates in K3, where UNKNOWN never counts as FALSE.

The word "store"

Merged docs disagree, and this PR does not settle it:

• The capstone calls DT64 "a versioned semantic binding-store adapter".
• D-DT64-Q0's "One-sentence contract" says "not the persistent Binding Store".

Decision 4 sides with the capstone. Plan §4 requests a Q0 amendment that defines "store" as bindings plus their version log, never copied business data. This PR does not apply it.

Staged work (each stage needs its own approval)

D-id	Deliverable
D-DT64-6-0	This plan and its board entries
D-DT64-6-1a	Classify all 523 verb files in both readers (today they accept 12 and 511, disjoint)
D-DT64-6-1	Reconcile the two readers' partitions over one checkout in one run
D-DT64-6-2	One (src, verb, dst) table from the entity ogit:allowed blocks plus the 12 from-to verbs; default source OgitTables::Allowed; no second parser
D-DT64-6-3	Register OGIT verbs as DT64 predicates in 64-wide blocks
D-DT64-6-4	Consumer read-only access through HotPlug activation
D-DT64-6-5	Demote OGIT to a cross-check oracle once a domain has source-attested predicates

Open questions are in plan §6.

Board changes

• Adds the board entry entries/2026-10-28-d-dt64-6-routing-bindstore-behavior.md.
• Prepends rows to INTEGRATION_PLANS.md (now eight decisions) and STATUS_BOARD.md.
• Regenerates entries/README.md.

Validation

• Every in-repo source is cited by symbol or section heading and was read in full. That includes ttl_parse.rs, OGAR ttl.rs, sgo.rs parse_verb, ogit.rs and tests/ontology_cache.rs. The OGIT counts are file counts, with their method recorded in the plan.
• Run locally against origin/main, all passing:
  • citation_decay.py --since origin/main (0 new decays);
  • entries_index.py --check;
  • supersession_index.py (no change).
• No code or tests were run, since this PR changes no code.
• REJECTED for failing task D-SAG-6-W0/W1: dogfood workbench amendment + read-only lab fixture #1812 is quarantined, and nothing here derives from it.

---

🤖 Generated with Claude Code

https://claude.ai/code/session_01XAqKMNiyjwqFke9CfmoBgn

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: 5afe8edb-a4b9-45e6-83be-17f09751c882

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-28T01:04:04.235192Z	5b37f8f	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[coderabbitai]

Add Carrot credits or activate Agent usage billing to use Autofix

[AdaWorldAPI]

Review of #1831 (head 73705ca)

The plan reads cleanly. I re-read three of its in-repo citations and all three hold: the verbatim Q0 line 9 quote, the HotPlug fields, and the ttl_parse module doc. The findings below are about the OGIT predicate expansion (D-DT64-6-1 to 6-3), because that is what DT64 needs first. Every number comes from a command listed at the end, run on OGIT master b5731a8 (2026-10-27), lance-graph main eeaaafb and OGAR main 6b43a48.

1. ttl_parse recognises 12 of OGIT's 523 verbs (blocks D-DT64-6-1 and 6-2)

classify in crates/lance-graph-ontology/src/ttl_parse.rs treats a subject as a verb only when it is a rdfs:Class; rdfs:subClassOf ogit:Verb. In OGIT, 511 of the 523 verb TTL files declare owl:ObjectProperty instead, and only the 12 NTO/WorkOrder/verbs/* files match that shape. The other 511 fall through to SubjectKind::Other and produce no edge proposal at all.

The plan describes the gap as "ogit:allowed is not yet expanded". It is wider than that: 98% of the verb vocabulary never reaches ttl_parse in the first place. D-DT64-6-1, as written, will report this as a partition mismatch. I'd add an explicit first step: classify owl:ObjectProperty verbs, or at least report them as unclassified, before any ogit:allowed work.

2. ogit:allowed lives on entity files, not verb files (D-DT64-6-2 targets the wrong place)

• Where the triples are. 960 TTL files carry ogit:allowed, and none of them are under verbs/. 940 are under entities/; the others are 12 under NTO/Credit/Entities (capital E), 7 under NTO/Healthcare/enumerations and 1 at the root. Each block reads [ verb target ] inside an entity, so the entity is the source, as in NTO/Schedule/entities/Calendar.ttl.
• The verb-level declaration is rare. ogit:from-to appears in only the same 12 WorkOrder verb files.
• What D-DT64-6-2 should do instead. D-DT64-6-2 proposes to "replace the placeholder LinkSpec", but ttl_parse builds that LinkSpec once per verb. The (src, verb, dst) rows have to come from the entity pass, plus the 12 from-to verbs. A per-verb LinkSpec could only ever be a union over all entities. ttl_parse's own module doc says "the detailed ogit:allowed constraint blocks for verbs", which carries the same misplacement, and is probably where it started.
• How OGAR already does it. OGAR reads ogit:allowed exactly this way: ogar-from-schema/src/ttl.rs extract_allowed_block fills EntityDecl.allowed: Vec<(String, String)>, and OgitTables turns that into Table::Allowed. That reference is missing from the plan's grounding table.

3. D-DT64-6-2 creates a second ogit:allowed parser

OGAR depends on lance-graph-ontology, so lance-graph cannot reuse ogar-from-schema. D-DT64-6-2 will therefore produce a second reader of the same block. The plan's guard is "No second registry". That is about stored output; it doesn't cover two parsers drifting apart. D-DT64-6-1 compares the two once. A durable version of that check would be a standing parity test in OGAR, the only side that can see both readers. crates/ogar-action-handler/tests/ontology_cache.rs already loads both under the ontology-cache feature, so the test could go there.

4. The 523 / 2,894 versus 518 / 2,756 gap

• Not OGIT drift. I counted OGIT at both 2315167 (the revision odoo-rs §8 measured) and master b5731a8. 24 TTL files changed between them, but every count I took is the same at both: 523 verb TTL files, 953 entity TTL files, 931 subClassOf ogit:Entity, 960 files with ogit:allowed, 12 with ogit:from-to. I did not count allowed edges, so this rules out drift only for these file-level counts. What is left points at reader differences (rejected, duplicate and unresolved entries, plus dedup rules), which is exactly what D-DT64-6-1 is for.
• The 518 / 2,756 figures have an in-repo source. The plan treats them as an unverified board claim. They are written in merged docs/research/epistemic-awareness/eaw-5-odoo-ogit-law-and-learning-reality-check.md line 24 ("939 types, 518 verbs, 2,756 allowed edges"), which can be cited. They still don't belong in the same column as 523 / 2,894.

5. Consistency

• Seven decisions or eight. The INTEGRATION_PLANS.md row says "Records seven user decisions" and lists seven. The plan and the board entry record eight; the row is missing D8, the register copy.
• Line-number citation. The same row, the board entry and plan §4 cite "Q0 line 9" by line number, while the plan's rule is to cite by symbol or section.

6. The crai-rs framing

• Successor or consumer. CR-H0 keeps D-SAG-6 as "the motor" and forbids a new Codegraph owner in crai-rs (CR-H0-CHASSIS-RETROFIT-MAP.md §7). The crai-rs ROADMAP makes CR-H2 wait on the D-SAG-6-1 audit. That makes crai-rs a sidecar that consumes docs(D-SAG-6): Rust Codegraph → Mathematics → Logic → Tracert, prompt and board #1760, not "the successor to docs(D-SAG-6): Rust Codegraph → Mathematics → Logic → Tracert, prompt and board #1760" as the plan and board entry call it.
• Stale status in CR-H0. CR-H0 §2 lists docs(D-CFA-0): counterfactual flow analysis capstone + thought dependency scratchpad #1813–docs(EAW-10): minimal legal-rule AST and conditional deciding-question gate #1830 as "currently open docs proposals" as of 2026-10-27. They are all merged now.

7. Smaller points

• Ruff role names. §7 says the ruff role names were not checked against ruff. The source is AdaWorldAPI/ruff crates/ruff_spo_triplet/src/role.rs, which the merged CFA thought-dependency scratchpad already cites.
• No Rust frontend in ruff. "Ruff harvests source structure" (§3) has a limit for Rust. ruff main f2cb523 has SPO frontends for C++, C#, Python, Ruby, MeTTa and SQLAlchemy, but none for Rust. For Rust code the transmission currently has no Ruff input.

Suggested order for the OGIT stages

1. D-DT64-6-1a: ttl_parse classifies, or reports, the 511 owl:ObjectProperty verbs.
2. D-DT64-6-1: run both readers and compare their partitions, as already planned.
3. D-DT64-6-2: build the (src, verb, dst) table from entity ogit:allowed blocks plus the 12 ogit:from-to verbs. Report unresolved targets rather than dropping them, and add the standing parity test in OGAR.
4. D-DT64-6-3: as planned.

Commands (OGIT checkout, master b5731a8)

git grep -l 'ogit:allowed' origin/master -- '*.ttl' | wc -l                        # 960
git grep -l 'ogit:allowed' origin/master -- '*.ttl' | grep -c '/verbs/'            # 0
git grep -l 'owl:ObjectProperty' origin/master -- '*/verbs/*.ttl' | wc -l          # 511
git grep -l 'subClassOf ogit:Verb' origin/master -- '*/verbs/*.ttl' | wc -l        # 12
git grep -l 'ogit:from-to' origin/master -- '*.ttl' | wc -l                        # 12
git grep -il 'subClassOf ogit:Entity' origin/master -- '*.ttl' | wc -l             # 931
git ls-tree -r --name-only origin/master | grep -c '/verbs/.*\.ttl$'               # 523
git diff --stat 2315167 origin/master -- '*.ttl' | tail -1                         # 24 files changed

I added no code and ran no tests; this review is reading and counting only.

---

Generated by Claude Code