💾 Archived View for jacksonchen666.com › posts › 2023-07-01 › 17-02-05 › index.gmi captured on 2024-03-21 at 15:24:23. Gemini links have been rewritten to link to archived content
⬅️ Previous capture (2023-11-04)
-=-=-=-=-=-=-
2023-07-01 15:02:05Z
The Fediverse is decentralized.
Cool, but what does that mean? And how does that actually work?
Well, compare that to a centralized site like Twitter or something: Everyone connects to *it*. And if it breaks, well, it's broken. Big F.
On the other hand, the Fediverse is a little bit different. Each Fedi instance is a server. Many users can live on one Fedi instance. The Fedi instance federates with the rest of the network, resulting in federated interactions and stuff. In this case, one instance going down is everyone on that instance being unable to use their thing, their Fedi instance.
But can we do better than one instance affecting potentially thousands of people? Yes, it's peer to peer (P2P). With P2P, one peer going down is one peer going down. If you have more than one peer, things are still going fine, just without that peer.
Now introducing another concept: The Client–server model. You connect to a central server, and get stuff from it. That's basically how the world wide web works.
However, there's a single point of failure with the Client–server model: The server. If it dies, you cannot look at Twitter posts or use your Fedi instance. Now, you can make the server redundant, but most Fedi instances might not do high availability because it could be quite costly and Twitter would like to destroy itself.
Twitter restricts viewing tweets to logged in users only
With P2P however, it is not a Client–server model. Peer-to-peer is where peers connect directly to each other. So instead of having central servers going down potentially affecting some or all users, one peer going down doesn't kill the entire thing (unless that's the only peer left, then it'll kill the entire thing until someone comes along).
That brings up the question: Is the Fediverse decentralized *and* centralized at the same time? Well, given that an instance going down can affect potentially up to thousands (or hundreds) of users, it is kinda like that.