Mastodon seems to be a lot more fragile than USENET; the whole PubSub thing seems to assume better connectivity than old-style distribution permits
This blog has been offline from ActivityPub / Fediverse for a few days, due to a bug in the WordPress Supercache plugin regarding (not) honouring the HTTP Accept header.
Back in the days of USENET, a node going offline for that length of time would not be a huge problem; the articles would be pushed or offered downstream to other peers, and dropped if they were deemed too stale (generally 1 week, often more) and all that would happen is a brief spike.
Yes, it’s a different world with different choices Alec, blah blah blah blah blah, but I am not impressed. If the ActivityPub architectures are not tuned in the expectations of delivering robustness of communication in the face of outages, it is going to (for instance) lose all account of protests in <some oppressive regime> where the in-country instances can only achieve intermittent connectivity in the face of state attempts to block communication.
It could be WordPress at fault, it could be Mastodon, it could be the ActivityPub spec or the recommended timeouts or who knows what cause; I cannot say and I am also not going to be the one to fix it because I also have bigger fish to fry.
But nor am I going to leave this issue unobserved. We need to talk about weakness in order to have it addressed.
