The difference is one uses W3C recommended web interfaces and the other does not. I make blog posts to my website using HTML, anyone can open them using a web browser. I can host those pages one of many static hosting services with no need for a special protocol or anything like that. AT on the other hand reïnvents this entire thing. Can you type the AT URI of a resource into the browser and have it open? I do not believe you can. So if I want to publish HTML content I now need both my regular web server and an AT PDS running on my server. There exists, as far as I'm aware, no application that can view HTML hosted on the AT network so I have to produce all of my blog posts in an agnostic format or publish them to no one, or I can use one of the services that automatically posts a link to that HTML as a BSky post. So you have all of that architecture and what it accomplishes is posting a link to content hosted elsewhere.
Why, instead of creating an indexer that works with existing formats, did they choose to create one intentionally incompatible with 99.999% of web content? Because in practice it's not a web protocol, it's not a serious thing. It's just a whole bunch of bullshit for nerds to make blog posts about that serves as the back end of a Twitter clone. It is standards proliferation. It adds nothing and fragments existing efforts. Instead of contributing to the web ecosystem, it subtracts from it. It breaks the important rules of software development, to do one thing and do it well, to use existing solutions. Everything it does is duplicated, there's a new way to sign documents, a new way to host content, a new data format, a new exchange protocol, a new way to serve documents, a new way to send notifications, etc. It's not a tenable way of doing things. If every person looking to make a web indexer created an entire separate web ecosystem, that would not be sustainable. They're clearly not intending to replace the Web, so what are they doing? It's just a complicated way to store tweets.
And the silliest part is that, if I recall correctly, they don't even have decentralised network yet since only they can issue PLC DIDs.
Something based on ActivityPub would be much more reasonable. I would not have nearly the same objections if it integrated with an existing ecosystem rather than carving up the already limited pool of people interested in the open web and locking a bunch of them into a different silo to the rest of us for essentially no technical advantage.
But ActivityPub works a certain way and it's different from how ATProto works. You're saying HTTP should've been a layer on top of email - how would that ever work?
Their primary goal was creating a Twitter clone, ActivityPub can do that. The federation stuff is really just a toy part of it, as you can tell from the fact that PLC DIDs are still issued from a central authority. Or they could have done it the way I suggested, with indexes of HTML pages. Both would be preferable.
Why, instead of creating an indexer that works with existing formats, did they choose to create one intentionally incompatible with 99.999% of web content? Because in practice it's not a web protocol, it's not a serious thing. It's just a whole bunch of bullshit for nerds to make blog posts about that serves as the back end of a Twitter clone. It is standards proliferation. It adds nothing and fragments existing efforts. Instead of contributing to the web ecosystem, it subtracts from it. It breaks the important rules of software development, to do one thing and do it well, to use existing solutions. Everything it does is duplicated, there's a new way to sign documents, a new way to host content, a new data format, a new exchange protocol, a new way to serve documents, a new way to send notifications, etc. It's not a tenable way of doing things. If every person looking to make a web indexer created an entire separate web ecosystem, that would not be sustainable. They're clearly not intending to replace the Web, so what are they doing? It's just a complicated way to store tweets.
And the silliest part is that, if I recall correctly, they don't even have decentralised network yet since only they can issue PLC DIDs.