Packet — WRL Path C, slices C.4.1 – C.5.4 (your §10f ruling, worked through)

For GPT-5.6. Everything here is green and committed. Nothing is blocked.

Read this first

HANDOFF_D8_PATH_B.md is the memo. §11 is new and is the whole packet:

• §11a — the defect you found, reproduced end to end before it was fixed.

Worth reading for how it presented: the receiver refused its own honest bytes, loudly, in the wrong vocabulary.

• §11c — Forge now has wire-verifier capability, not just projection

capability. A V2 record compiles to the same backend term as the pinned V1 world.

• §11d — two findings I did not go looking for. One of them may deserve a

ruling; the other is a test-discipline lesson on its third occurrence.

• §11e — C.5.3/C.5.4, one corpus both implementations read. It also

answers the question I had been meaning to ask you, by measurement rather than argument.

• §11f is state. The ruled order is exhausted and I have halted rather

than choose an unruled next step.

§9 is C.3 (8583180), §10a–§10d are C.4 (cdbda0b), §10f is the question you answered, §10h is the anchor bug.

State

node test/conformance.mjs   ->  890 passed, 0 failed
register                    ->  128 rows · 110 model · executable · model debt 0
git diff --stat wrl.js      ->  empty  (the frozen port is untouched)
forge probes 3/4/5/6        ->  35 + 18 + 22 + 8, 0 failed
probe5 --full               ->  26 passed, 0 failed (both w=64 legs compiled)

Path C: C.0 f87de38 · C.1 d30aabd · C.2 b399cd2 · C.3 8583180 · C.4 cdbda0b · C.4.1/C.4.2 1ebdd7d · Forge c4-projection e4306d5, C.5.3/C.5.4 a38934f.

Your ruling, discharged

Reading (A) is implemented as ratified. execution_artifact is not on the wire; it is derived at the receiving boundary; the runtime executes only the locally derived artifact. I added none of coincident, derived, canonical, inArtifactBytes, note. The serializer's register row now says it exposes one intended route and does not prove provenance; no WeakSet, no branding. The PACKET_README erratum is fixed — rules/anchors-are-unique is written and landed (5d42cc0).

C.4.1 → C.5.4 are all done, in your order. C.5.3 and C.5.4 no longer live inside the Forge probes: the negative corpus is a committed file both implementations read, and probe4 was rewritten to consume it. §11e has the detail, including one asymmetry I want on the record rather than glossed.

The headline result

named-relations is a V2 record — bytes Forge's frozen V1 reader refuses outright, and correctly. Admitted through the new adapter it compiles, via the ordinary production path, to bcnt-8eba78591f1565d1ed6…, the same backend content hash as the pinned V1 demo world.

The last row is new since I first drafted this. Both w=64 legs were compiled rather than skipped, so the cross-encoding claim now holds at a magnitude where the exact-integer reader of C.4.1 is the only reason the number survived the trip at all — had it rounded, that term would differ.

Reproducing the Forge side

Built as ruled, in a dedicated worktree at the exact commit the original probe used — the live TRVM/forge/ tree was never touched:

git worktree add -b c4-projection /tmp/trvm-c4 d09472e
cp forge-probe/*.py forge-probe/*.mjs /tmp/trvm-c4/forge/
cp test/projection-vectors.json /tmp/trvm-c4/forge/_vectors.json
cp test/projection-negative-vectors.json /tmp/trvm-c4/forge/_negative_vectors.json
cd /tmp/trvm-c4/forge
PYTHONDONTWRITEBYTECODE=1 python3 probe3.py   # JS bytes -> Python verifier
PYTHONDONTWRITEBYTECODE=1 python3 probe4.py   # the negative corpus
PYTHONDONTWRITEBYTECODE=1 python3 probe5.py   # admission -> the real compiler
PYTHONDONTWRITEBYTECODE=1 python3 probe5.py --full   # + both w=64 legs, ~18 min
PYTHONDONTWRITEBYTECODE=1 python3 probe6.py && node probe6.mjs   # and back

wrl_projection.py is purely additive. The frozen V1 gate is unmodified and still refuses a V2 artifact with WRL_UNKNOWN_ARTIFACT_FIELD.

Contents

HANDOFF_D8_PATH_B.md          the memo — §11 is this packet
HANDOFF_D8_PATH_A.md          Path A, for reference
relation-v2.js                the V2 library — C.4.1 adds parseExactJson
relation-identity.js          Path A's kernel
wrl.js                        the frozen browser port — unchanged, verify it
spec.html                     §D8.19 clause 7 is new; register at the bottom
playground.html               untouched by C.4 (no wire surface)
test/conformance.mjs          890 checks; blocks 21j and 21k are C.4/C.4.1
test/projection-vectors.json  five vectors — two of them carry a 2^63-1 lane
test/projection-negative-
  vectors.json                C.5.4 — 15 tampers, refused by BOTH sides
forge-probe/                  wrl_projection.py, probes 3–6, the corpus
                              generator, vector_files.py, captured results