[jerojero]

"There are 4 standards that try to solve this problem, its too many, we need one that finally unifies it all and solves the problem once and for all" "There are 5 standards that..."

Jokes aside, I think we need stronger arguments as to why something like activity pub is not good enough to solve the problem instead of trying to come up a new way of solving the "decentralized comms" problem.

[danabramov]

ActivityPub and atproto are differently shaped. Pitting them against each other is like asking “why need web when we have email”.

ActivityPub is email-shaped. Servers are inboxes sending messages to each other.

atproto is web-shaped. User repositories host data (like personal sites or git/RSS), while apps aggregate from repositories (like Google Reader).

Different topologies lead to different properties. Eg atproto lets user change hosting with no disruption in app experience. atproto also lets anyone build new apps aggregating over existing data.

ActivityPub doesn’t allow either of those things. It’s literally a bunch of small centralized coupled hosting+app services messaging each other.

[class4behavior]

Calling AP services a bunch of small "centralized" services in this context removes all the meaning from that term. You might as well call any web server centralized while comparing them to clouds.

Proper federation is exactly such bunch of small services messaging each other. On the hand, what ATProto leads to is at most a handful of large-scale providers each running the own portion of the network.

[danabramov]

There’s a clear difference in architecture between

1) a layer of app-agnostic hosting providers + a separate independent layer of apps aggregating over data from those (like personal sites with RSS + aggregators like Google Reader)

2) a circle of flat instances where each node couples app+hosting (like many little Twitters)

One doesn’t couple hosting with apps, another one does.

Mastodon/AP model is (2), atproto model is (1). You should be able to see the outcomes from different network shapes.

In atproto, you can build a new app that works with existing data, but in AP you can’t. In atproto you can move hosting with zero effect on your identity or how you show up in apps, in AP you can’t.

[knowtheory]

I dunno man. Why was Tangled able to ship on top of ATProto even prior to getting funded, and ForgeFed has been hanging out for years?

[Kye]

That's become my answer to all "why not ActivityPub?" questions.

AP isn't completely stagnant but there's a reason AT is still holding on to and accelerating that early developer excitement AP had. Maybe it's marketing, maybe it's money, maybe it's some technical thing. Maybe it's the community. Whatever it is, people seem to enjoy developing in the Atmosphere in a way I never saw on AP.

[aniviacat]

If you consider people's worries of centralization happening with AT, that's not necessarily a good sign.

Of course VC funding and startup mindsets prefer a protocol that is easier to rugpull.

[nerdypepper]

its linked in the original post as well, but here is an explanation of why activitypub is not a good fit for this problem, by the authors of ForgeFed themselves: https://forgefed.org/blog/actor-programming/

[compyman]

Reading that - I'm really not sure that AT Protocol has a much better story there either.

(as I understand it) the data has to live in a PDS, PDS are keyed by accounts, so you are similarly stymied for collaborative projects? I guess AT Proto is still a real work in progress so maybe that story has improved since the last time I checked it out.

[nerdypepper]

> But federated authorization is one of the things ActivityPub doesn't define, and leaves it to us to figure out.

this is the key bit, atproto has this. sidecar services like knot can use service authentication[0] for authenticated requests.

[0]: https://atproto.com/guides/auth

[pfraze]

Yeah the problems they seemed to have were over collaborative data structures with permissions. You’re right about how atproto solves that, which means you’re using CRDTs if you need to collaborate. If that’s a fit mismatch, I’d tell people to just appoint api servers which wrap a repo and provide the needed semantics.

[knowtheory]

Yeah, capability for group permissions is a key part of the work happening on permissioned data in ATproto right now.

https://dholms.leaflet.pub/3meluqcwky22a

https://dholms.leaflet.pub/3mfrsbcn2gk2a

https://dholms.leaflet.pub/3mguviy6iks2a

https://dholms.leaflet.pub/3mhj6bcqats2o

[tardedmeme]

Or email. AP is very similar to SMTP over HTTP.