# P9 reproducibility record

Status: internally accepted paper artifact; P9 runtime remains `OBSTRUCTED`

Date: 23 July 2026

## Evidence boundary

This manuscript is a source-level research artifact. A reader can reproduce
its LaTeX build and audit its citations, evidence labels, proposed semantics,
falsification conditions, and explicit nonclaims.

The current boundary is:

| Layer | Present evidence | Not established |
|---|---|---|
| P9 evidence dossier | source-grounded prior-art boundary, 13 claim anchors, 99 proposed fixture cases, 10 proposed Lean targets, and falsifiers | accepted workspace profile, implementation, experiment, or theorem |
| Existing CatDB repository | bounded schema, path, and mapping work owned by prerequisite papers | workspace commit, overlay, semantic diff, merge, rollback, or historical-query runtime |
| P9 manuscript | definitions, conflict taxonomy, implementation order, evaluation protocol, and exact nonclaims | executed workspace result, performance result, or machine-checked P9 law |

Every CatDB-specific P9 claim remains `OPEN`, `OBSTRUCTED`, `PROPOSED`,
`ENGINEERING HYPOTHESIS`, or `OPEN CONJECTURE`. Prior-art statements are the
only claims supported by existing literature in this author pass.

## Clean build

Run the build in a fresh temporary directory so no PDF or auxiliary file is
written beside the manuscript:

```sh
build_dir="$(mktemp -d /tmp/catdb-p9-final.XXXXXX)"
cp paper.tex references.bib "$build_dir/"
cd "$build_dir"
pdflatex -interaction=nonstopmode -halt-on-error paper.tex
bibtex paper
pdflatex -interaction=nonstopmode -halt-on-error paper.tex
pdflatex -interaction=nonstopmode -halt-on-error paper.tex
pdfinfo paper.pdf
```

Acceptance checks:

```sh
rg -n \
  "Overfull|undefined references|undefined citations|LaTeX Error|Emergency stop|Fatal error" \
  "$build_dir/paper.log"
rg -n "duplicate ignored" "$build_dir/paper.log"
rg -n "Warning|error|not found" "$build_dir/paper.blg"
texcount -inc -sum "$build_dir/paper.tex"
```

The author pass also converts the PDF to one PNG per page and inspects every
rendered page:

```sh
render_dir="$(mktemp -d /tmp/catdb-p9-final-render.XXXXXX)"
pdftoppm -png -r 72 "$build_dir/paper.pdf" "$render_dir/page"
```

## Current corrected-build observation

The final isolated build for the source hashes below is:

- temporary build directory: `/tmp/catdb-p09-r1fix.Iqde06`;
- format: 12pt article, US letter, PDF 1.7;
- page count: 49;
- PDF size: 626,747 bytes;
- TeXcount: 9,269 text words and 9,887 summed words;
- bibliography entries: 29;
- overfull boxes: 0;
- undefined references: 0;
- undefined citations: 0;
- BibTeX warnings: 0;
- duplicate PDF destinations: 0;
- LaTeX errors: 0; and
- underfull boxes: 5.

The five underfull boxes occur in a repository-status paragraph, two narrow
technical paragraphs, one appendix table cell, and one bibliography entry.
Rendered inspection found no clipping or overlap from them.

All 49 corrected pages were rendered to images and inspected in a 7-by-7
contact sheet. The author pass had separately inspected the title page,
implementation diagram, reproduction schema, and final bibliography page.
No heading, text block, equation, table, figure, citation, link, page number,
or sidebar was clipped or overlapped.

Temporary build hashes:

- `paper.pdf` SHA-256:
  `fe35315a91ce5fc561d8e2032fcc72cf391a067db1054dfe65bbafe19575d875`;
- `paper.log` SHA-256:
  `5da0b7c54860f719f6aa3c44caf162ebd0728e4f15f73877e426d150d4a31742`.

Repository source hashes:

- `paper.tex` SHA-256:
  `78afb7850fabc5f1373c0f62affb6cc939f545c756b6b0622616b0ac7111da03`;
- `references.bib` SHA-256:
  `fdfcc72262cb061a43386d9c4eef8334df5d5b6d296b826ebb3cf72a02d7ee2e`;
- evidence dossier SHA-256:
  `fbd5cd02b8802a82c0b7845ceade32890043bc1e419b9a71acf0a3a5c8a18e2f`;
  and
- bounded TeX-plus-bibliography source digest:
  `8e454075bb4190db4f3529dae102444c57e1a24d1028a067dcc610b4644e2716`.
- repository `paper.pdf` SHA-256:
  `fe35315a91ce5fc561d8e2032fcc72cf391a067db1054dfe65bbafe19575d875`;
- repository `paper.log` SHA-256:
  `5da0b7c54860f719f6aa3c44caf162ebd0728e4f15f73877e426d150d4a31742`;
- repository 300-DPI `cover.png` SHA-256:
  `161715d44e3edb03606874e0b9440ee58e8e2fed84d91d315b7dab2608ed738c`;
  and
- canonical AGY review SHA-256:
  `33fb0079943eb2edf4483560ebab2f8b63a6037e7f8b167e76509f116895e854`.

The repository retains the accepted PDF, inspected log, and 300-DPI
first-page cover. It does not retain BibTeX, table-of-contents, or other TeX
auxiliary build files.

## Paper review

AGY round 1 reviewed combined source digest
`b4de7031c6e16492c0743fae250754b5c9e368e8d330832d284642ead1d6130f`
and requested an external-compensation receipt contract plus a concrete
non-pushout merge example. Round 2 reviewed the corrected digest
`8e454075bb4190db4f3529dae102444c57e1a24d1028a067dcc610b4644e2716`,
verified both corrections, found no remaining blocker, major, or minor issue,
and ended `VERDICT: ACCEPT`.

## Citation closure

The final BibTeX pass emits 29 bibliography entries. The build has no missing
or undefined citation. The paper uses the primary-source set from the P9
evidence dossier, including database versioning, schema and mapping evolution,
model management, categorical data migration, historical query, data
citation, provenance, persistent data structures, MVCC, and materialized-view
adaptation.

Citation verification for the evidence dossier independently resolved:

- 20 Crossref DOI records;
- 2 DataCite DOI records; and
- the arXiv, W3C, IBM, and VLDB primary records used by the dossier.

This verifies bibliographic identity and source relevance. It does not promote
any CatDB workspace claim.

## Future semantic reproduction contract

A future P9 result bundle must retain:

1. exact repository revision and dirty-state record;
2. frozen workspace, artifact, operation, normalization, conflict, authority,
   provenance, retention, and reproduction profiles;
3. immutable input artifacts and content digests;
4. base, target, and source commit graphs;
5. hand-derived fixture expectations and manifest hashes;
6. independent Rust and Haskell normalized outputs;
7. Lean declaration names, axiom inventory, and placeholder audit where a
   machine-checked result is claimed;
8. complete dependency, diff, impact, merge, validation, reconciliation, and
   publication traces;
9. crash and concurrency failpoint logs;
10. historical-query envelopes, normalized data digests, and provenance
    digests;
11. baseline source, reconstruction notes, exclusions, hardware, dependency
    versions, seeds, raw metrics, failures, and timeouts;
12. reviewer decisions tied to exact artifact digests; and
13. a retained record for every rejected, unsupported, or unresolved case.

An implementation that reads expected fixture outputs as its result is not an
independent oracle. A successful reconstruction is not automatically an exact
query replay, and equal data without equal required provenance is not exact
result reproduction.

## Deliberate absences

This author pass did not:

- implement P9 code, fixtures, benchmarks, or Lean declarations;
- run Rust, Haskell, Lean, SQLite failpoint, PostgreSQL, distributed, or user
  experiments;
- invoke AGY or create paper-review artifacts during the author pass; AGY
  round 1 was invoked later and is preserved separately;
- promote an `OPEN` or `OBSTRUCTED` claim;
- create a repository PDF, cover image, log, or auxiliary file during the
  author pass; the accepted PDF, cover, and log were added only after AGY
  acceptance;
- commit or publish changes;
- edit `README.md`, `CITATION.cff`, shared coordination files, code, Slack, or
  another paper; or
- claim that Git text diff supplies semantic database merge.

Independent domain, systems, adversarial, code, reproducibility, and
human-readable review remains a later paper-review stage.