Shadow evaluation — not public

Japan Desk

Two official Japanese-language RSS feeds published by the Ministry of Defense: its news stream and its site-update stream. Not an English press-release source, and not the Joint Staff, neither of which either feed carries.

Shadow testing, scheduled daily and publishing nothing. RSS discovery works: two official Japanese-language ministry feeds answer normally and support conditional requests. Body retrieval is partial. Documents published as PDF are served and are collected in full; documents published as HTML are returned behind an interactive challenge and are kept as titled, dated discovery records with no body text and nothing inferred. Measured across the news feed on 2026-08-26, 8 of 142 items (6 percent) carried full text and 134 of 142 (94 percent) did not. That is partial retrieval, not coverage, and it does not establish complete Ministry of Defense or Joint Staff coverage.

Records: None collected
Sources configured: 4
Sources enabled: 0

This desk has collected nothing into the record. No count on this page describes material this project holds.

Sources

Source	Institution	Authority	Enabled	Records
防衛省・自衛隊 — news feed (Japanese)	—	—	No	none
防衛省・自衛隊 — site update feed (Japanese)	—	—	No	none
統合幕僚監部 Joint Staff — press releases (English)	—	—	No	none
Ministry of Defense — press releases (English)	—	—	No	none

An enabled source is one the pipeline is configured to visit. It is not a promise that a working collector exists for it — see Coverage, where each source's own outcome is recorded every run.

Pre-registered source universe

The source universe is published before collection begins, so the declared scope can be checked against what is eventually collected. This is a statement of intent and of researched fact. It is not coverage, and nothing below has produced a record.

Sources identified

Listing and item format verified directly; no endpoint is claimed that was not checked.
Source	Item format	Note
Ministry of Defense — news feed (Japanese)	RSS — 142 items measured 2026-08-26: 134 HTML, 8 PDF	Families: /j/press/ 105, /j/approach/ 37.
Ministry of Defense — site update feed (Japanese)	RSS — 391 items measured 2026-08-26: 342 HTML, 0 PDF	A whole-site stream, not a press-release feed: /j/press/ 126, /j/budget/ 93, /j/profile/ 76, /j/approach/ 63, /j/policy/ 21, /j/presiding/ 11, /j/kids/ 1. Every item is labeled by its family and nothing is filtered out.
Joint Staff — press releases (English)	HTML index — challenged (403), not collected	Declared so the Joint Staff stays distinguishable. Carried by neither RSS feed, and its own index is challenged. Nothing has been collected from it.
Ministry of Defense — press releases (English)	HTML index — challenged (403), not collected	Declared so the English estate stays distinguishable. Nothing has been collected from it.

Observed publication volume

Joint Staff press releases — 2026, to mid-August: 135
Joint Staff press releases — 2025, full year: 214
Joint Staff press releases — 2014-2026, whole archive: 895

These figures describe what the ministry published and what its feeds carry. They are not a record count and nothing here has been collected into the public record. Only documents published as PDF carry text; on the news feed that was 8 of 142 items on 2026-08-26.

What has to be solved before this desk can collect

Documents published as HTML cannot be retrieved.: The ministry's RSS feeds and its PDF documents are served normally, but HTML documents on the same host are returned behind an interactive challenge. Those items are discovered and recorded as unretrieved rather than inferred to be absent. An institution that wants to refuse this collector must be able to recognise it and say so, so defeating the challenge would destroy the capability that rule protects. Resolving this means requesting an official route.
Collection-health thresholds remain unset.: Publication cadence and silence thresholds are calibrated only after shadow collection has run. Guessing now would create false alarms or hide real silence.

What has already been solved

Recorded rather than deleted. A blocker that quietly disappears from this page leaves a reader unable to tell whether it was solved or abandoned.

Official discovery access — resolved 2026-08-26.: The listing pages remain challenged, but the ministry publishes two official RSS feeds that are served normally and support conditional requests. Discovery no longer depends on the challenged pages.
PDF text extraction wired to a collector — resolved 2026-08-26.: The dormant extractor is now used by the Japan shadow adapter, and by nothing on the production collection path.
Repeated titles — resolved 2026-08-26.: On the 2026 listing “Japan-U.S. Bilateral Exercise” appears 27 times for distinct events, and a title-only rule would collapse 26 of those 27 records. Deduplication is by canonical URL and by content hash of the extracted text, never by title, so a recurring ministry title stays as many records as there were publications.

Limits of this desk, stated plainly

No source is enabled in production. Nothing collected here reaches the production database, the public site, or any public count.
Retrieval is partial, and most sampled bodies are challenged. Measured on 2026-08-26, 8 of 142 news-feed items (6 percent) carried full text; the other 134 are stored as titled, dated discovery records with no body. This is not complete Ministry of Defense or Joint Staff coverage.
Both collected feeds are Japanese-language. The English estate and the Joint Staff are declared but not collected, and neither is reachable.
Shadow testing began 2026-08-26. No qualification period has elapsed and no desk-readiness claim follows from it.

These are properties of the desk a reader needs before citing it — see Coverage for what each collection run actually returned.

← All desks