新鲜度定义

Freshness answers one narrow question: when was this configuration last observed in any source we fetch? It is a property of our observations, not of the server. A "fresh" entry was seen in a recently fetched public document; an "archived" entry has not been seen for a long time.

How buckets are computed

Whenever a source is fetched successfully, every entry recognized in the document updates its last_seen_at timestamp. The bucket is then a pure function of now - last_seen_at:

• fresh — last observed within the last 3 days (72 hours).
• recent — last observed 3 to 14 days ago.
• aging — last observed 14 to 45 days ago.
• stale — last observed 45 to 90 days ago.
• archived — not observed for more than 90 days. Archived records are kept for historical

reference and trend analysis, are excluded from default listings and searches, and can be filtered out of API queries and personal feeds.

Buckets are recomputed on ingest and on read; they never depend on client-side caching.

Reappearance

If an entry that had fallen to *stale* or *archived* shows up again in a fetched source, its last_seen_at resets and the bucket improves accordingly. The timeseries counts such events as configs_reappeared. A single reappearance is treated exactly like any other observation — no special trust is attached to entries that come back.

What freshness is not

• Freshness is not an uptime or reachability measure. Nobody at HiddenIO connects to the listed

servers; a fresh entry can already be offline.

• Freshness is not a safety or quality signal. Malicious or honeypot entries can be published

continuously and therefore look fresh forever.

• Freshness is not guaranteed to be complete. We only see the sources we fetch; a configuration

that quietly persists outside our source list will age out even if its server still runs.

FAQ

Why 3/14/45/90 days? These breakpoints match the publish cadence we observe in practice: most public lists rotate within days, weekly lists within about two weeks, and anything unrefreshed for a quarter is usually abandoned.

Can I filter by freshness? Yes — in the search filters, in the API (freshness=fresh,recent), in subscription specs and in personal feeds.

Why keep archived records at all? Keeping them lets the changelog and statistics distinguish "we stopped seeing it" from "it never existed", and lets researchers study how public configurations appear and disappear. Archived records are always explicitly labeled.

Do observations update freshness for all sources of a record? Yes — a record merges observations from every source that contains it, so last_seen_at reflects the most recent fetch of *any* covering source.