This video from @jgarr about the differences between #ActivityPub and #ATProto has me wondering if ActivityPub could actually support separating identity from content, since things like post ids are fully qualified URLs.

Could I host my identity on one domain and my content on another? I haven’t found anything in the protocol spec that forbids that, but I may have overlooked it.

https://youtu.be/wJBCpzM1VfM

ActivityPub vs AT Protocol

YouTube
@jamie That's an interesting question; I've been wondering how reasonable it would be to introduce content stores shared by multiple hosts to reduce storage needs and possibly offer other advantages. Sort of a backend CDN.
@trevorbramble IIUC this may be another problem ATProto is trying to solve through what they call Data Repositories. https://atproto.com/guides/data-repos
Personal Data Repositories | AT Protocol

@jamie Ah-ha. I was surprised not to see even a bad early attempt at solving it yet, but frankly I've been ignoring Bluesky. (And Threads.)

@trevorbramble I've mostly been ignoring them, too, but I do like some of the ideas behind ATProto. Portable identity, owning your data, stuff like that.

The data is in Merkle trees, though, which sounds like it contains an immutable history. If that's accurate, I don't like it because, to me, ownership of my data means I can delete things as if they never existed.

@jamie Oh, hum. Portable identity is on my short list of vital features.

The case for showing edits is because jokers will change their post to make it look like everyone agreed that cannibalism is the best diet, or something, right? (And maybe official tweets from park rangers or reporters can't be invisibly defaced by ne'er-do-wells?)

@trevorbramble lol Not sure. I assumed it was for data provenance at a lower layer, which has broader implications. If everything is immutable at that layer, people can also tell what you posted and deleted, liked and then unliked, etc. Basically everything becomes a git commit.