Post

The non-technical version: Every person gets their own database that stores their own likes, comments, etc. You register those individual databases with the Bluesky server who then goes and asks for your data from your database and compiles it with everyone else’s to create your feed.

I was waiting for this Neo! 🚀 Directly to my weekly readings Thanks for sharing!

Awesome breakdown, @systemdesign.one! Love how you simplified Bluesky's federated model and ATProto framework. The use of an index server for processing streams is a smart approach for scaling. Excited to check out the visuals in your newsletter—thanks for sharing this insightful thread!

• Bluesky is a decentralized social network; it distributes control across servers. • They built Bluesky on top of a decentralized open-source framework called ATProto. • They use a federated architecture for its familiar client-server model, and convenience.

• They define separate data schemas and API endpoints for apps built on ATProto. • They store data published by a single user in a repository; it uses SQLite as the data layer. • The repository stores information about every action by a user such as a like, or comment on a user's post.

• They store repositories of different users on a data server; it exposes HTTP for clients. • They use the crawler service to collect information from each repository. • The crawler forwards the collected data as a stream over websockets.

That's an awesome outline, Neo!