AMR-050-0062: portable exact-checker packaging recommendation
===========================================================

This is a packaging inspection, not a new mathematical execution or proof.
check_focal_pair.py has not been changed or rerun for this inspection.

Runtime and minimum material
----------------------------
The only required executable is check_focal_pair.py. Its imports are Python
standard-library modules: argparse, datetime, fractions, functools, hashlib,
json, math, and pathlib. It contains its own exact quadratic-field arithmetic,
orbit-incidence/reflection/caustic checks, focal inversion, and shoelace code.
There are no repository imports or external packages. References to earlier
AMR-050-0044/0046 helpers are methodological credits, not runtime dependencies.
Python 3.9 or newer is required by functools.cache and Path.is_relative_to.

Frozen checker SHA-256:
4d231ca868f086eb99da882b9c90b3ef391bb99f33af29c3324819a0ac9170b1

A compact public source-archive layout is:

  reproducibility/check_focal_pair.py
  reproducibility/README.txt
  reproducibility/exact_controls.txt
  reproducibility/originating-exact-result.json

The originating result can be a byte-exact copy of exact-run-001.json.
exact-isolated-run-002.json may additionally be included as a prior isolated-
interpreter receipt. Neither receipt substitutes for a new extracted-archive
replay. The executable itself needs none of the documentation or result files.
The general proof belongs in the manuscript, separately from finite controls.

Suggested README usage
----------------------
From a freshly extracted source-archive root, run:

  python3 -I -B reproducibility/check_focal_pair.py

To save an additional result, choose a path that does not already exist:

  python3 -I -B reproducibility/check_focal_pair.py --output reproducibility/replayed-result.json

The output path must be inside the directory containing the checker. A checker
placed under reproducibility/ therefore does not accept an archive-root
result.json. Existing output files are deliberately not overwritten. The
checker otherwise writes no files and prints its JSON result to stdout. It
does not use network access, credentials, installation, or repository state.
-I isolates Python from user/repository import paths; -B avoids bytecode files.
Do not use -O or PYTHONOPTIMIZE: mathematical checks use Python assertions.

The expected successful status is:

  PASS_EXACT_FOCAL_PAIR_PRODUCT_CONTROLS

Recommended package-stage verification
--------------------------------------
The publisher should create the actual archive, extract it into a fresh
temporary directory, check the extracted checker SHA-256 above, and run the
README command there using an isolated interpreter. Save the actual exit code,
stdout/result, extracted script hash, and archive hash in the package receipt.
This procedure is recommended, not claimed to have occurred in this inspection.

The previously retained exact-isolated-run-002.json is PASS from an isolated
interpreter on the original source path, not a freshly extracted ZIP. Its
repository_dependencies is [], floating_point_used is false, and
finite_controls_are_general_proof is false. Prior exact runs agreed in semantic
content after removing actual_execution_completed_at. A new replay must also
exclude that truthful execution timestamp when comparing results; byte-exact
JSON equality would incorrectly demand the same execution time. Keep the
checker bytes unchanged so script_sha256 also agrees.

Scope and limitations for the public README
------------------------------------------
This executable verifies finitely many exact controls, not the general odd-N
invariance theorem. It checks two non-equivalent genuine primitive triangular
phases of one fixed ellipse/confocal-caustic family, odd repetitions of those
triangles, orientation/reindexing controls, a circular control, and a genuine
even-N negative control. A passing run does not prove all odd periods, does not
constitute a proof-assistant formalization, independent-human review, or a
novelty certificate. There is no exact noncircular primitive-five or star-orbit
fixture in this checker. Any general result must be supported by the separately
stated manuscript proof rather than extrapolated from these finite tests.

Optional numerical diagnostic: keep separate
--------------------------------------------
scout_focal_pair.py and numerical-run-001.json concern an optional ordinary
85-digit mpmath diagnostic. They are neither exact nor interval-certified, and
are not proof. The diagnostic's reported small residuals/product spreads only
record sampled behavior, including primitive odd star families.

scout_focal_pair.py currently obtains mpmath from the repository's local
tmp/aim0014_python_deps path. It is therefore not portable as-is in an isolated
source archive. Do not include it in the required standard-library checker
test list or copy that dependency directory into the public archive. A compact
package may omit the scout executable and retain its diagnostic result with
explicit provenance. If executable diagnostic reproduction is wanted later,
use a separately authored optional portable adapter with an explicitly stated
mpmath dependency. No such adaptation is claimed here.
