Adopted RFC-0001 RDF source alignment
Versioned draft snapshot
RFCs, reference catalogs, and evidence reflect Spec 3b2e3e7a114fb826b1445d39a48756b8fff629cd. Counts and pending/status notes belong to that snapshot. Read the dated project update for later progress.
Source repositories are private during staging. Links marked “repository access required” need access to the linked repository.
This is a trace and gap record from the pinned draft specification. It does not establish proof or implementation conformance.
Earlier implemented existing-contract corrections
Two of the fifteen remaining RFC-0001 semantic mismatches were routine source omissions, rather than choices of a new contract:
- The
action.backcompatdomain summary called the authoring intent itself Phase-1. It now permits Q2 authoring and defers only prior-version validation, matching authored §5, the existing matching claim nodes, and Lean diagnostic and tradeoff catalogs. - Lean statement
b54omitted the authored §6 requirement that violations of the no-silent-coercion rule surface diagnostics. That clause already existed in the RDF claim and Lean anti-pattern catalog; the statement now includes it.
source-only.patch is the exact applied two-source-line correction.
actual/alignment.{textproto,pb} is the single grouped record of original
fifteen-node dispositions, source identities, actual commands, generated
changes, review preservation, and query comparisons. Its twelve sibling logs
record twelve successful generation/integrity commands, including compilation
of all 55 structured Lean RFC modules. All 38 actual query result multisets are
unchanged. The semantic delta replaces one domain summary and one emitted RFC
statement predicate; 56 other emitted Turtle catalogs are byte-identical.
All 138 RFC-0001 projection reviews bind the new Lean source hash. Exactly two
verdicts change to matching; every unrelated finding and verdict is preserved.
At that checkpoint the global ledger retained 2,418 mismatches, including thirteen RFC-0001
identity/kind/class decisions. Only b54’s citation hash changes in the provenBy
ledger; its mismatch finding, theorem, and verdict remain intact. Diagnostic
definitions and coverage are unchanged.
The initial controller failed before any generator or native command when it
used a dictionary with a repeated protobuf message field. Its source and exact
four-file effects were retained under the original path named in the typed
precommand_failure record. Only those owned files were restored to the base
commit, and the metadata construction was repaired with explicit message
add() calls. The twelve recorded commands then executed once. That failure
record is a labeled derivative of the actual tool response, not a fabricated
process log. The private historical controller is not bundled here.
Reviewed public RDF decision
The maintained claim-model proposal
consolidates the thirteen decisions. The owner subsequently approved the complete
batch in reply call359cf486; its actual adoption is recorded below.
pending-owner-review.patch: compact exact seven-file owner-choice patch, verified to produce byte-identical proposed payloads to the measured candidate.pending-owner-impact.{textproto,pb}: complete candidate’s 38-query comparison and explicit query column contracts.pending-complete-candidate.patchandpending-complete-impact.{textproto,pb}: original complete candidate and its actual measurements. The original diff used repeated-line heuristics that produced unnecessarily large textproto hunks;prepare_review_patch.pyverifies the compact patch against all seven proposed file payloads without rerunning queries or other spec gates.pending-identity-candidate.patchandpending-impact.{textproto,pb}: retained initial five-source-file candidate and measured impact before the necessary query-consumer corrections were added.proposed-queries/: exact reviewed query sources, now adopted canonically.pending.proto,pending_consumer.proto, and the preparation Python sources describe the measured counterfactuals and preserve input hashes.
The candidate replaces 28 RDF triples and adds 71: the graph grows from 89,301 to 89,344 triples. Six of 38 query result multisets change. The two proposed query corrections make domain entities without DSL FQNs visible as named IRIs and classify orphan subjects by declared domain type. These are explicit public query-contract changes that the owner approved. The initial impact retains the observed false-orphan rows rather than hiding that issue.
Actual approved adoption
approved-actual/adoption.{textproto,pb} binds the exact approved patch,
source base, human approval reference, before/after sources, actual commands,
review changes, RDF delta, and all 38 query comparisons. adopt_approved.py
applied that patch once and invoked the owning projection generator and
integrity gates. Seventeen existing review hashes change; thirteen mismatch
verdicts are corrected; three review identities move; five new matching reviews
are added. Unrelated findings and verdicts are preserved. The result is
6,959 reviewed nodes, 2,405 mismatches, no unreviewed nodes, and no remaining
RFC-0001 projection mismatches.
Actual RDF removes 28 triples and adds 71. All 38 actual query deltas exactly match the reviewed counterfactual, including the six changed result multisets. Nine of ten invoked existing integrity commands pass. The strict domain-identity gate returns 1 on inherited conflicts rather than being relabeled successful.
approved-actual/checks.{textproto,pb} proves that complete finding set is
identical to the source base: 18 FQN and 21 kind-class conflicts. It also binds
the finite retained 24-jar Jena 5.0.0 closure, JDK 21 binaries, two unchanged
existing Java sources, and canonical RDF inputs. The Java sources were compiled
into an owned private directory; live SHACL conforms and the existing malformed
document fixture is rejected. The four Java version/compile/run commands all
return 0. No framework or Roma build, package download, source-copy edit, or
dependency-pin change was used for this local validator run.
Evidence limits
Records preserve their actual original paths, before/after hashes, process
statuses, and comparison scope. They do not bundle every source input or tool
runtime and are not a standalone reproduction environment. align.py mutates
maintained files and must not be rerun to inspect the evidence. The approved
adoption helper also mutates maintained files and must not be rerun for review.
No parser conformance, proof soundness, dependency pin, source-copy cutover, site update, publication, or final clean package is established. The genuine generation receipt precedes the final documentation updates; final committed artifact generation requires a fresh receipt.