Bridgy Fed (via granary) includes https://purl.archive.org/miscellany in the @context of every activity it delivers. The URL serves the SWICG "ActivityPub Miscellaneous Terms" context and is hosted on purl.archive.org, which in our experience is unreliable: over the last 90 days we've observed failures resolving this context on 28 separate days, typically in multi-hour bursts
When purl.archive.org is unreachable, Activity.fromJsonLd() fails with jsonld.InvalidUrl ("loading remote context failed") for every Bridgy Fed originated activity, before any inbox listener runs - the same failure mode that motivated preloading Lemmy's context (#714) and http://joinmastodon.org/ns (#630). The default kvCache TTL of 5 minutes
offers no protection against outages longer than that
For context: on our deployment (Ghost's ActivityPub service), purl.archive.org outages cause these failures every few days - most recently for several hours on both Jul 19 and Jul 20 - each time producing sender retry storms and 5xx alerting noise
The context seems like a good preload candidate:
- It's a SWICG deliverable, deliberately stable and versioned, with an immutable URL for the current version
(https://purl.archive.org/miscellany/1.0.0)
- It's small (four term definitions:
Hashtag, manuallyApprovesFollowers, movedTo, sensitive) all aliasing AS2 terms
- Fedify already preloads from the same unreliable host:
https://purl.archive.org/socialweb/webfinger
- Fedify's own vocab docs reference the spec (
hashtag.yaml links to https://swicg.github.io/miscellany/#Hashtag)
I have a PR ready (same shape as the Lemmy change in #714: JSON asset + preloadedContexts entry). Will open it referencing this issue
Bridgy Fed (via granary) includes
https://purl.archive.org/miscellanyin the@contextof every activity it delivers. The URL serves the SWICG "ActivityPub Miscellaneous Terms" context and is hosted onpurl.archive.org, which in our experience is unreliable: over the last 90 days we've observed failures resolving this context on 28 separate days, typically in multi-hour burstsWhen
purl.archive.orgis unreachable,Activity.fromJsonLd()fails withjsonld.InvalidUrl("loading remote context failed") for every Bridgy Fed originated activity, before any inbox listener runs - the same failure mode that motivated preloading Lemmy's context (#714) andhttp://joinmastodon.org/ns(#630). The defaultkvCacheTTL of 5 minutesoffers no protection against outages longer than that
For context: on our deployment (Ghost's ActivityPub service),
purl.archive.orgoutages cause these failures every few days - most recently for several hours on both Jul 19 and Jul 20 - each time producing sender retry storms and5xxalerting noiseThe context seems like a good preload candidate:
(https://purl.archive.org/miscellany/1.0.0)
Hashtag,manuallyApprovesFollowers,movedTo,sensitive) all aliasing AS2 termshttps://purl.archive.org/socialweb/webfingerhashtag.yamllinks to https://swicg.github.io/miscellany/#Hashtag)I have a PR ready (same shape as the Lemmy change in #714: JSON asset +
preloadedContextsentry). Will open it referencing this issue