basically I am pretty skeptical the relay design is gonna be able to support many independent instances
ActivityPub as-deployed has many challenges, if you've read enough of my writing you'll certainly hear that (and some of how to improve the situation) but at least it started under a foundation of a lot of nodes participating on the network
BTW re: "content which survives AP nodes going down", I made a demo of how to do that in an AP-spec compatible way a few years ago, and it even had encryption so not everyone hosting the content needed to know what it was https://gitlab.com/spritely/golem/blob/master/README.org
Also I spent some time reviewing Bluesky's DID stuff enough to be fairly convinced that it won't, in a useful way to users, really survive nodes going down in the ways users believe (that's on top of did:plc being centralized, which is well known, and did:web being completely superfluous, which seems obvious but I'm surprised I don't see advocated as often). I have a lot of analysis on this stuff, but I'm hesitant to post some of this because I just sound like I'm being grouchy.
But really, I'm worried about users being sold on the idea of decentralization and it falling out from under their feet and losing faith in it, when there's architectural challenges there from the get-go. That's why this really bugs me.
Maybe they can be addressed, and that's one reason to post the critique, but I suspect it'll just make people mad at me, I dunno
@stefan AP currently does not survive nodes going down well. That's addressable, and I have written out part of the answer of how to address it, but it's true that it doesn't do that presently and I'm not sure there's community interest to fix it.
I actually think that the people working at Bluesky do believe in decentralization, and they have a lot of good people there. If there are ways to collaborate which are fruitful, I'd rather not burn a bridge, which is partly why I keep my mouth shut. But there are serious, serious structural problems for achieving decentralization and I don't think the financial model will prioritize fixing them, and in the meanwhile many users mistakenly believe they *are* on a decentralized system.
I very much agree with @pluralistic's concerns that bsky need to *actually ship* meaningful, non-reversible federation (but also trust the folks involved to do that. There will be a *lot* of side-eyeing and loss of faith in progressive movements if they don't 😅).
I've wanted to write up a "critique document" of both "present day ActivityPub" and ATproto but wasn't sure how they'd be received. I think knowing you'd be interested makes me feel pretty positive about it
I do think that @blaine is right in that there's shall we say a "fixed point of convergence" if things were to work right, and so maybe that's the right way to frame things around