Live data from Hacker News

The AT protocol is the most obtuse crock of shit

urbanists.social

11–20 of 493 posts

Re: The AT protocol is the most obtuse crock of shit

#15

Any other embedded programmers or older people that remember what an AT protocol really is confused every time they see a story about Bluesky?

With the first comment mentioning XRPC, I was even more confused, since that's the name of a protocol from around the same time period. I believe TI-RPC won that battle.

Re: The AT protocol is the most obtuse crock of shit

#16
post #8
post #3

> It turns out using Git, which is almost always used with a centralized 'remote', to do federation, which needs to be weakly consistent, IS A BAD IDEA!!!!! Given that a user owns their feed, I'm not sure why this is a bad idea? I want to take this critiques here seriously but there's not a ton to grasp onto. It definitely didn't feel great that AtProto built from bare ground, reinvented json-schema & OpenAPI for no…

My limited understanding of the problem about git as the basis for user data storage is that it then would expose the complexity of git rebasing and the rest via federation. Git is great with a centralized remote (like I believe the original post is getting at) but less so in a federated setting.

Rebasing and similar operations are only a hassle because multiple authors can edit the same files, so you end up with conflicts.

In the scenario where every authors can only change their own files, you can avoid those sort of conflicts, as they'll never happen.

Remaining is how to solve files referencing other files (in the case of a federated social media, posts that are replies to other posts) which are way easier to solve, as it's not really conflicts but invariants that have to be created.

Post reply on HN