🧵 [1 / 7]
When #Bluesky says "Protocols, not platforms", they intent two things:
1. Grabbing people's attention by telling them what they want to hear.
2. Presenting the AT Protocol as an alternative to #ActivityPub to capitalize on the current hype around #Mastodon, the #Fediverse and decentralized social media in general.
The two protocols are not equal solutions for the same problem and, in fact, AT is not even a (communication) protocol to begin with.
🧵 [2 / 7]
The purpose of (communication) protocols is to standardize how two separate systems can talk to each other. They are interfaces that allow for different implementations.
ActivityPub [1] is such an interface, describing how to accept and forward content in a federated network. There are many implementations of the protocol, each serving a different application: Mastodon, Friendica, PixelFed,...
All can talk to each other and are inter-operable.
🧵 [3 / 7]
AT isn't exactly a protocol. The scope of its specification largely describes an implementation, not an interface [2] . Granted, that implementation is very flexible and allows for different applications to be build on top of it, so it may look like a protocol, but it misses the crucial part (on purpose): vendor independency.
🧵 [4 / 7]
It is not (realistically) possible to build an AT based system without relying on Bluesky code and/or their network, which puts Bluesky in the position to enforce rules and, when push comes to shove, step in to sabotage competitors that are getting too big [3] . The whole point of a standardization is to prevent exactly that.
🧵 [5 / 7]
The problem, AT (Authenticated Transfer) solves has nothing to do with federation or decentralization. It's a method for moving a user's content around in a cloud (or between clouds) and updating the storagelocation/userhandle relationship in a directory afterwards.
The purpose is to allow Bluesky to rent server capacity, on demand, from the big cloud providers and jump between their datacenters. It's simply part of their strategy of avoiding getting vendor locked-in themselves.
🧵 [6 / 7]
In the broader sense, an authenticated transfer of content might be called a protocol, depending on the definition of the term [4] . And that is what Bluesky is playing at here: putting a retrofit component of their system in the spotlight, labeling it "protocol" and using the ambiguity of the term to break into the public discussion about open communication protocols with a protocol for database synchronization, for the purpose of promoting its own service.
Sources:
I honestly think the issue here may be that documentation hasn't been finished, that AT is a communication protocol, just one that's not fully documented yet.
It's plenty legit to criticize devs for getting behind on documentation, but that is a different criticism.
Being fully documented is the prerequisite for calling it a protocol. If it is not, then a third party wouldn't know how to talk to system, thus defeating the point of a communications protocol.
Ha! As if proprietary protocols aren't a thing?
Yes, a protocol is a protocol even if you yourself haven't been assisted in joining the party yet.
