# Paper 10 Reproducibility Record

- **Manuscript:** *Distributed CatDB: Typed Operation Algebra and Coordination Boundaries*
- **Recorded:** 2026-07-23
- **Repository revision inspected:** `ec3b7d3d13f946b70fd74acd0d0b75d8924f97fa`
- **Result:** PASS for manuscript compilation, rendering, exact-digest AGY
  review, and repository artifact verification; NOT RUN for a distributed
  CatDB experiment.

## Evidence boundary

This paper is a provisional semantics and engineering-contract paper. The
repository inspection found no:

- `catdb-distribution` crate;
- typed P10 operation log;
- causal-context or logical-clock implementation;
- replicated conflict resolver;
- partition or healing simulator;
- convergence fixture;
- checkpoint or membership protocol;
- independent distributed Haskell oracle;
- distributed Lean model;
- backend crash-recovery harness; or
- P10 benchmark result.

P10 also depends directly on unresolved P9 workspace and P13 provenance
interfaces. Consequently:

- no CatDB convergence, recovery, availability, latency, throughput, or
  metadata-overhead result is reported;
- the 25-operation matrix is an `ENGINEERING HYPOTHESIS`;
- unimplemented bounded components remain `OPEN`;
- runtime, recovery, and performance claims remain `OBSTRUCTED`;
- the twelve counterexamples are paper-level results under their displayed
  finite definitions; and
- external CRDT, causal, database, consensus, and schema-change results remain
  prior art rather than CatDB evidence.

The frozen manuscript and bibliography later passed a paper-specific AGY
review bound to the exact combined source digest. The canonical review is
byte-identical to round 1 and ends in `VERDICT: ACCEPT`.

## Authoring inputs

The manuscript was derived from the paper-local evidence dossier and checked
against the shared P10 manifest, dependency graph, terminology, roadmap, and
benchmark-plan boundaries without editing those shared files.

| Input or source | SHA-256 |
|---|---|
| `research/evidence-dossier.md` | `48a706afb433c8348290f219629b30b24b8c86a3fcb1e49b85dc9578005e04e8` |
| `paper.tex` | `08112d7f3fc3820f269292ac8ebbcab13b52c0caf1fd34a12ea9f6a5b3806172` |
| `references.bib` | `2026f43b9d04e281bbf22ecbef155dd4cab0ced84c102184ca5be491f4baa6c7` |

The bibliography contains 29 primary-source records. The final `.bbl`
contains 29 `\bibitem` entries. BibTeX emitted zero warnings, and the final
LaTeX log contains zero undefined citations.

## Clean isolated build

The final verification began in a newly created directory containing only
`paper.tex` and `references.bib`:

```text
/tmp/catdb-p10-accept.kjR0Ly
```

Equivalent reproduction commands from the repository root are:

```sh
build_dir=$(mktemp -d /tmp/catdb-p10-final.XXXXXX)
cp papers/10-distributed-catdb/paper.tex \
   papers/10-distributed-catdb/references.bib \
   "$build_dir"/
latexmk -cd -pdf -interaction=nonstopmode -halt-on-error \
   "$build_dir/paper.tex"
```

`latexmk` began without auxiliary files, ran pdfLaTeX, ran BibTeX, and repeated
pdfLaTeX until the table of contents, cross-references, citations, and
bibliography stabilized.

### Toolchain

| Tool | Version |
|---|---|
| latexmk | 4.88, 9 March 2026 |
| pdfTeX | 3.141592653-2.6-1.40.29, TeX Live 2026 |
| BibTeX | 0.99e, TeX Live 2026 |

### Final temporary outputs

| Artifact | SHA-256 |
|---|---|
| `paper.pdf` | `ebfeb8dd7c6a492f78d97530b886fea585e0c4d8bd5f428c1ed6bb073ffa012a` |
| `paper.log` | `5f3a757d7bfd5361048e21210bfbd82562f54d97fbcb7f965c4106d2476439aa` |
| `paper.bbl` | `00bfa2ef9b649a421cb637ba3900d27f2e0d9e351dbacde1c6632c14986018b4` |
| `paper.blg` | `983c99aaf0a6f5dc09565f79ef6582ada95b7cf89cc52b6d8897ad3c326c2afc` |

The temporary PDF is PDF 1.7, 377,660 bytes, and contains 38 letter-sized
pages. A binary PDF hash can vary across rebuilds because PDF metadata and
trailer identifiers may encode build-time state. The manuscript, bibliography,
and generated `.bbl` identify the content.

## Warning inventory

| Check | Count | Disposition |
|---|---:|---|
| TeX errors | 0 | pass |
| Undefined citations or references | 0 | pass |
| Multiply defined labels | 0 | pass |
| BibTeX warnings | 0 | pass |
| Overfull boxes | 0 | pass |
| Underfull boxes | 0 | pass |
| `everypage` legacy-package notices | 2 | accepted formatting-package notice |

The `everypage` notices arise from the arXiv-style first-page GrokRxiv sidebar.
They do not alter page content or citation resolution.

## All-page render inspection

The final PDF was rasterized at 110 DPI:

```sh
mkdir -p /tmp/catdb-p10-accept.kjR0Ly/rendered
pdftoppm -png -r 110 \
  /tmp/catdb-p10-accept.kjR0Ly/paper.pdf \
  /tmp/catdb-p10-accept.kjR0Ly/rendered/page
```

The 38 page images were assembled into ordered four-page sheets and inspected:

| Sheet | Pages | Surfaces checked |
|---|---|---|
| 00 | 1--4 | title, vertical sidebar, abstract, TOC, introduction |
| 01 | 5--8 | contributions, evidence tables, prior art, citations |
| 02 | 9--12 | running example, listings, definitions, equations |
| 03 | 13--16 | replay proposition, CRDT contract, clocks, matrix start |
| 04 | 17--20 | matrix continuation, schema/mapping listings, CAP corollary |
| 05 | 21--24 | coordination table and all twelve counterexample surfaces |
| 06 | 25--28 | replay listings, checkpoint model, Lean and Rust roadmaps |
| 07 | 29--32 | commands, baselines, falsification list, limitations |
| 08 | 33--36 | conclusion, all 28 nonclaims, claim ledger, references start |
| 09 | 37--38 | bibliography completion |

Page 28 was also inspected independently at full raster size because adjacent
page numbers were visually close in the contact sheet.

Inspection found:

- no clipped text or page number;
- no GrokRxiv sidebar collision;
- no table, equation, listing, or proof overlap;
- no malformed longtable continuation;
- no text outside the page boundary;
- no broken hyperlink text;
- no missing bibliography entry; and
- no blank or duplicated page.

ImageMagick emitted a nonfatal empty-font warning while assembling unlabeled
contact sheets. All ten sheets were produced, and the warning did not affect
the underlying page rasters or their inspection.

## Source and style checks

Final source checks found:

- 38 pages, exceeding the 20-page minimum;
- 25 operation-matrix rows;
- twelve numbered counterexamples;
- 28 exact nonclaims;
- 22 claim-ledger entries;
- 29 bibliography entries;
- no Unicode em dash or en dash;
- no curly quotation marks;
- no unresolved authoring placeholder; and
- the accepted repository PDF, log, and 300-DPI first-page cover are exact
  copies of the final acceptance build outputs.

The humanizer scope and source hashes are recorded separately in
`humanizer-scope.md`.

## Accepted repository artifacts

| Artifact | SHA-256 |
|---|---|
| combined `paper.tex` plus `references.bib` source | `879d9e326c60439786d09c54b21b0733e79de94a01e86935be72f4be9e8b8f52` |
| repository `paper.pdf` | `ebfeb8dd7c6a492f78d97530b886fea585e0c4d8bd5f428c1ed6bb073ffa012a` |
| repository `paper.log` | `5f3a757d7bfd5361048e21210bfbd82562f54d97fbcb7f965c4106d2476439aa` |
| repository `cover.png` | `5f4f0cbd6db9c294f2a504478b485ed51ac1352f615352a44e1debd17c1efac9` |
| canonical AGY review | `c51da9a0522295726722e87408dc277d87a6778daec70a7a5e665ade278a796c` |

The final repository verifier confirms source-digest identity, PDF freshness,
38 pages, a clean LaTeX log, cover freshness, 300-by-300 DPI cover metadata,
the exact review digest, and terminal `VERDICT: ACCEPT`.

## Current reproducibility claim

What is reproducible now is the manuscript and its internal evidence
classification. A reader can rebuild the same 38-page content from the two
paper-local publication sources using the toolchain above. The build resolves
all citations and references, and the rendered pages show the complete
operation matrix, counterexamples, nonclaims, and bibliography without layout
defects.

No distributed CatDB behavior is reproducible because no such implementation
or fixture exists in the inspected repository.

## Required future execution bundle

Every future P10 run should archive:

```text
benchmarks/runs/<run_id>/
  pre-registration.md
  source.json
  build.json
  environment.json
  baseline-manifest.json
  history-schema.json
  histories/
  delivery-schedules/
  raw-observations/
  canonical-states/
  conflicts/
  minimized-failures/
  statistics.json
  provenance.json
```

The bundle must preserve exact source revisions, seeds, membership epochs,
typed operations, delivery schedules, partitions, healing, duplicates, crash
points, durability boundaries, step results, canonical states, conflict sets,
and minimized failures. Correctness must pass before performance is reported.
Lean proofs, Haskell properties, Rust histories, and backend recovery runs must
retain their distinct evidence classes.