Source archive layer
This directory documents how the database’s claims could be re-checked after the original pages, videos, or builds disappear. It contains receipts, not copies: the Internet Archive stores the page content; this database stores the snapshot URL and the time it was requested.
Files
| File | Role |
|---|---|
source-urls.tsv |
The URLs eligible for snapshotting: 31 videos, the Miro board, and any manually added press/official pages. |
source-archive.tsv |
Receipts: one row per requested URL with the Wayback snapshot URL, HTTP status, and notes. |
source-manifest.tsv |
One row per event (610), classifying what the claim rests on and whether a web snapshot can help. |
sources/README.md |
This policy document. |
What is archived today
- All 31 YouTube videos. These are the primary record for everything in the database. If a video is deleted or made private, the snapshot preserves the title, description, and whatever YouTube rendered at capture time. All 31 snapshots were re-checked and returned HTTP 200.
- The Miro board. A capture was recorded, but Miro refuses archived replay
(HTTP 403), so the snapshot is marked
blocked. the board export remains the durable copy of its contents. - Nothing else yet. The event summaries rarely record URLs for the press
articles, patch notes, store pages, or datamine posts they cite. Those rows are
marked
unlocatedinsource-manifest.tsvrather than given invented links.
Current counts and the list of blocked or failed rows: sources/archive-status.md.
What a web archive cannot do
Most of the corpus is build evidence: strings, VData schemas, FGD entries, shaders, entity names, and patch diffs. Wayback cannot archive a Steam depot. Verifying those claims requires:
- the game build ID and date,
- the specific file path or string token,
- a local, unmodified copy of that build or a trustworthy diff, and
- a note of what changed in the next build.
Do not redistribute Valve game files. Record the build ID, file path, and observation; keep any extracted evidence minimal and factual.
source-manifest.tsv separates the two worlds:
evidence_type=build→ needs a build snapshot.evidence_type=press,official,resume,community→ Wayback-eligible if a URL is recorded.evidence_type=other→ Tyler’s inference or uncited analysis; no artifact to archive.
Privacy
Casting and résumé evidence must never include an individual’s name, photo, or
profile URL. Archive the fact of a listing only if it can be redacted; otherwise
record “was public on index.md §8.3, item 3.