Coverage

What was collected, what was not, and which kind of "nothing" each source returned. This page exists because an empty result has more than one cause, and only one of them is healthy.

Only desks that collect appear in a run. A desk that is planned, researched, access-blocked or under shadow evaluation contributes no rows here, and its absence is explained on Desks rather than shown as a failure.

Most recent run

Run 122 · 2026-08-26 13:33:24 → 2026-08-26 13:36:50 · aggregate failed

Every configured source has a result row for this run. A missing row would be a defect. The desk shown is the source's declared desk.
Source	Desk	Result	Found	Parsed	Text read	Already held	New	Note
Parsed counts pages the parser returned a document for. Text read counts the documents that actually carried readable text. They differ when a page was reachable but unreadable: the item's title, link and date are kept, and its text is not. Runs recorded before this measurement existed read not measured rather than a zero they never observed.
china_mil_online	China Desk	Collected	8	8	not measured	0	6	Reached the listing and stored new items.
global_times_mil	China Desk	Collected	1	1	not measured	0	1	Reached the listing and stored new items.
mod_china	China Desk	Collected	9	9	not measured	8	1	Reached the listing and stored new items.
pla_daily	China Desk	Collected	34	34	not measured	4	27	Reached the listing and stored new items.
xinhua_mil	China Desk	Not implemented	0	0	not measured	0	0	Configured, but no working collector exists.

What each result means

Every outcome the pipeline can store has a plain-language meaning here. They are translated, not merged: two results that both produce zero records can mean opposite things, and the whole point of keeping seventeen of them apart is that they are seventeen different facts.

Result	Meaning
Collected ok	Reached the listing and stored new items.
Nothing published ok_no_publications	Reached the listing; it carried nothing new. This is healthy.
Nothing new ok_all_duplicates	Reached the listing; everything it offered was already in the archive. This is healthy.
Nothing relevant ok_all_filtered	Collected fine; nothing met the topic filter.
Not implemented not_implemented	Configured, but no working collector exists yet. It contributes nothing by design — and is reported every run so it cannot be mistaken for a healthy source.
Disabled skipped_disabled	Configured but switched off.
Could not reach listing_failure	The listing could not be retrieved or parsed. The empty result is a fault, not silence.
Access refused auth_failure	The source refused access — a login wall, or a refusal returned to an honestly identified collector. Nothing was collected, and nothing was done to defeat it.
Challenge required access_challenged	The publisher's edge asked for a browser challenge instead of serving the document. The site is up and the request was well formed — this collector simply is not being served that page. The item is recorded so the gap is visible; the challenge is not solved, retried, or worked around.
Timed out timeout	The request ran past its time limit. Whether the source had anything to offer is unknown.
Redirected away disallowed_redirect	The request was redirected outside the source's declared domains and was not followed.
Unexpected format unexpected_content_type	The response was not the document type the source declares. A challenge page served in place of a listing arrives this way.
Refused as oversized oversized_response	The response exceeded the size ceiling and was refused unread. No partial content was kept.
Could not retrieve documents fetch_failure	The listing was reached, but the documents it named could not be fetched. Items exist that we do not hold.
Could not read extraction_failure	Pages were fetched but nothing usable was parsed — usually the source changed its markup.
Collected, not analyzed analysis_failure	Collection succeeded and the records are stored; the analysis step failed for them. The record is intact.
Collector error adapter_error	The collector raised an unexpected error. Deliberately kept distinct from the specific failures above so an unclassified crash is never absorbed as one of them.
Unknown source unknown_source	A source was requested that no desk manifest declares. This is a configuration fault, not silence.

Access refusals are respected, not worked around

When a source refuses an honestly identified collector — a login wall, a redirect off its own domains, or a challenge page served in place of a listing — the refusal is recorded as the outcome and collection stops there. No browser impersonation, no challenge solving, no proxying. An institution that wants to refuse this collector must be able to recognise it and say so, and defeating a refusal would destroy the capability that rule protects.

Gaps in collection

Days with no collection run at all. These are outages, not quiet days — no attempt was made, so nothing can be said about what those days held.

From	To	Days
2026-07-17	2026-07-24	8
2026-08-21	2026-08-22	2

Why this table is on the site rather than in a changelog

A corpus with undisclosed holes cannot be cited safely. Anyone using this archive to argue that an institution went quiet during one of these windows would be wrong, and would have no way of knowing.