Oddbean new post about | logout
▲ ▼
tf | 3 months ago (raw) | root | parent | reply | flag 
 Sure it's from

https://bsky.social/about/blog/5-5-2023-federation-architecture
▲ ▼
Dr. Hax | 3 months ago (raw) | root | parent | reply | flag 
 I'm not sure which components are redundant here, but my guess would be the AppView because:

"An App View is the piece that actually assembles your feed and all the other data you see in the app"

So it seems that if the App View is down, you don't see your feed.

Of course, if your PDS is down you won't get any new messages either.

If a PDS can subscribe to multiple App Views, and it will catch up after it has been down (as opposed to just missing those messages/events), then that doesn't seem too bad.

Feed generators and labelers appear to be optional, though not much is said about them.
▲ ▼
tf | 3 months ago (raw) | root | parent | reply | flag 
 The way I see it (I'm thinking along the lines of deplatforming more than outages)...

Theoretically the AppViews are replaceable (assuming they're open source)

If the relay blocks your PDS and you connect to another relay, you need the entire rest of the architecture to also connect to that other relay to maintain your apps and frens

Each relay ingests the entire network

Each component ingests the entire network from each relay

So that's double the resources per component for 99.9999999% of the same data

Who is going to pay for that 🤔

I think it would have to be the new relay provider

Then where is the demand coming from to cover those costs