Bind P104 outgoing ports to recipient histories and input admission

[kumagi]

Change and mathematical scope

PR3423 left the actual histories behind its 675 outgoing ADD ports open. This source-bound audit closes those histories in the unchanged PR2586 word and identifies the additional entrance conditions. All port-cut states require 2,907 events/263 initial banks; preserving the terminal histories of all introduced dependencies stabilizes at 4,027 events/265 inputs (including private cleanup).

Under the explicitly inherited native covariance and Newton-module premises, every retained atomic state admits the sharp independent dyadic product entrance: 104 Y inputs in 4B9, 38 N inputs in 2B9, and 123 others in B9. At event 312, N100(x)=2*x_0 satisfies the earlier 63-bank admission but its exact integer half leaves B9. The complete terminal Y map has rank 9, certified by a factorization of all 52×104 entries and a minor of determinant -1/128. The inherited 2032H payload bound is specialized, not claimed as new.

This is a conditional interface certificate, not a paid supplier or κ improvement. It explicitly lists 696 omitted outgoing ADDs to 93 omitted banks; the projection does not preserve those side effects. Native semantics, input preparation, physical storage/routing/temporaries and recursive costs remain obligations. The selected frontier and unchanged-profile 599/584 stop are unchanged.

Task and overlap

• Base main: 6f100d4bb5083f8c35d96fd95a539811c62f245c.
• Task p104-side-correction, scope changed to the released PR3423 handoff; exact predecessor 15d28c4de37dc47ea07cd7173c8b318fca7a4d75.
• Canonical claim. Entry checks covered 353 open PRs and 2,277 Discussion records with complete pagination; publication refresh found no competing recipient-closure artifact. Unpublished work and exhaustive historical diffs are excluded.
• Complements Certify actual P104 Newton producer and consumer interfaces #3423 (open ports), Research: certify the sharp whole-word P104 Newton entrance #3406 (whole-word product domain), Research: independently certify all P104 native chart identities #3416 (native covariance), and Research: prove the correlated P104 ideal entrance #3426 (correlated final-cut admission). Runtime uses only the two unchanged source files already on main, not unmerged predecessor executables.

Validation

Python 3.12, standard library only. At 886222c1d3e2c2d878b263dc8d165ef03d5bac7f:

python3.12 -B scripts/research_package.py --package research/p104-external-port-closure
python3.12 -I -O -B research/p104-external-port-closure/verify.py
python3.12 -B scripts/check_research_commit.py --package research/p104-external-port-closure --check-source-history

All PASS. The 11-file source-bound archive passed export/check and isolated optimized replay with Git unavailable. Separate SSA and reverse-liveness implementations agree on exact closure sets. Checks include 142 weakened-admission controls, eight rejection groups, two rank regressions, exact rank witnesses, three full-word/projection scalar fixtures, and exact Newton negative/positive controls. A review-found rank-routine error was corrected before publication and recorded in TASK.json. Finite checks support the conditional proof in PROOF.md; no heavy native replay, Lean check, or selected-frontier replay was run.

Credit

Hiroki Kumazaki, with substantial OpenAI Codex analysis, code, execution, writing and separate AI-agent review; no independent human or proof-kernel review claimed. Apache-2.0. Credit Clear1nsight (#3423/#3382), ywz65 (#2586), Arham Tahir (#2665/#2674), Tekalign (#3025), Ross Richardson (#3068), Nick Petalas (#3406/#3416/#3426), Jose Manuel/Vasek's stopping audit, and the retained geometry/native/compiler lineage. Detailed provenance is in README.md and SOURCES.json.

Open next questions

The package's next questions start with supplying and paying for the 265-bank admission, or accounting for the 696 omitted side effects. Neither justifies another unchanged-profile supplier round; any changed profile first needs a compatible full cost screen.