I've been doing some research on #bluesky, nothing too deep. It piques my interest that they're trying to stand up an entire new federated protocol rather than just use (or even extend) ActivityPub. Not only does this make them incompatible with the Fediverse, it stands them in opposition to it with their own Fediverse-of-one. Remains to be seen if they'll be the VHS or the Betamax of this war.
The problem is that ActivityPub has some real deficiencies built into its core, that can't be simply extended away.
To address those issues--ranging from privacy through the focus on instances over users--requires an incompatible protocol.
@volkris as a developer I appreciate a green-field project as much as anyone, but to throw out the entire AP protocol rather than address issues (even with some adapter or glue code) strikes me as a very odd decision unless some other factor comes into play. It seems to me like an excuse to do what they wanted to do anyway.
Ha, well, there are two sides to this, objective and subjective. And please read the following in the lighthearted tone I mean it.
Objectively, it would take so much work to add in the features missing from ActivityPub, the user-empowering stuff, because AP is just that focused to the core on instances.
Subjectively, my personal opinion, AP is so ugly, so badly designed, so bloated and inefficient in its use of layer after layer of kludged-in standard, that I'm happy to throw it overboard.
I say this to respond to your strikes me as odd statement. To me, it doesn't strike me as odd at all, as from my perspective the alternative is a lot of work to maybe have new features sort of function in a system that's not work keeping in the first place.
And now to go take a vehicle to a mechanic to try to salvage a car that, if I'm being honest, is really not work keeping in the first place :)
So to be a bit more concrete, when I originally read the ActivityPub standard, it struck me as having a tremendous amount of overhead and being based on a design that would not scale well.
That's just looking at the technical side. There are other major issues, for example the lack of any guarantee of limited audience for a post.
But when I read the standard I wondered if the people behind it had any experience analyzing distributed systems for scalability, or if they'd even heard of Big-O analysis. (A professional programmer friend of mine says it's not necessarily even taught in college these days).
Sure enough, as the flight from Twitter took off, I heard case after case of instances literally crashing under the strain of those poorly designed protocols.
These issues are core to the AP design. You can't switch to lighter-weight signaling or more scalability-optimized distribution approaches without ripping it all out.
I'm hoping Bluesky hasn't made the same mistakes.
@volkris Oof. That's one savage code review. I understand though; sometimes something is just that broken. Or it's a round peg and you've got a square hole. Either way, the juice is not worth the squeeze.