The basic building block of the Fediverse is the instance, right? Every instance is its own self-contained, centralized social media service that optionally interacts with other instances. EG Trump’s Truth Social is a Mastodon instance that does not federate.
ATProto takes a more radical approach. Everything is modular. There is no instance or anything that is complete in itself. It’s more like the WWW. You can make websites in different ways. These are made findable through search engines like Google or Bing, which are not affiliated with companies offering web hosting.
ATProto takes everything apart. It tries to avoid choke points or lock-in as far as possible to thwart monopolies. You have a server that stores your data (posts, etc …), called a PDS. You can move your data to a different server. An identity provider tells others where your account is at any moment. A relay collects all the posts that people make and makes them available for further processing. This can be used to create algorithmic feeds, or moderation (aka labelling). These things are independent of each other and can be independently offered by different parties. You can pick and chose which to use, though there isn’t a whole lot of choice yet.
ETA: No idea what W wants to offer in that regard.
Follow-up question : doesn’t ATProto run into the same issue as Bluesky for example, where even though the protocol allows for decentralization, the vast majority of users are on the same servers, thus negating the pros of decentralization?
I feel like this would be the case here where one server will contain all the modules.
So, my main takeaway is that ATProto basically splits user authentication and content storage and feed algorithm? Like, these are 3 separate things?
You can have your own authentication server at home that basically just says that you are you (similar to a cryptographic identity), then you can store your content on google cloud and then use a 3rd-party feed algorithm?
The basic building block of the Fediverse is the instance, right? Every instance is its own self-contained, centralized social media service that optionally interacts with other instances. EG Trump’s Truth Social is a Mastodon instance that does not federate.
ATProto takes a more radical approach. Everything is modular. There is no instance or anything that is complete in itself. It’s more like the WWW. You can make websites in different ways. These are made findable through search engines like Google or Bing, which are not affiliated with companies offering web hosting.
ATProto takes everything apart. It tries to avoid choke points or lock-in as far as possible to thwart monopolies. You have a server that stores your data (posts, etc …), called a PDS. You can move your data to a different server. An identity provider tells others where your account is at any moment. A relay collects all the posts that people make and makes them available for further processing. This can be used to create algorithmic feeds, or moderation (aka labelling). These things are independent of each other and can be independently offered by different parties. You can pick and chose which to use, though there isn’t a whole lot of choice yet.
ETA: No idea what W wants to offer in that regard.
Thanks for taking the time to explain.
Follow-up question : doesn’t ATProto run into the same issue as Bluesky for example, where even though the protocol allows for decentralization, the vast majority of users are on the same servers, thus negating the pros of decentralization?
I feel like this would be the case here where one server will contain all the modules.
Bluesky itself is behind ATProto
Make of that what you will
So, my main takeaway is that ATProto basically splits user authentication and content storage and feed algorithm? Like, these are 3 separate things?
You can have your own authentication server at home that basically just says that you are you (similar to a cryptographic identity), then you can store your content on google cloud and then use a 3rd-party feed algorithm?