Conversation
Notices
-
Embed this notice
Hi @alex , how do you think nostr will scale when a user has to be connected to a big number of relays if he follows many accounts spread over those relays? How should it handle historic posts as pleroma does? Or is this what dito solves?
-
Embed this notice
@gvs It's almost the same issue on ActivityPub. Nostr is probably even a little better.
I'm more concerned about specific events in Nostr, like the way we store followings. But the core protocol itself I'm not too worried about.
-
Embed this notice
@gvs You have it right, but an ActivityPub server has the same problem. Every new server your friends create and join, your server must be aware of. Nostr isn't creating a new scaling problem, it's just shifting it to the client, which is good for freedom. Also, due to the flexible identity system there are much fewer relays than AP servers. And you can create relays that aggregate other relays, or hybrid servers like Ditto that serve the client a cache.
-
Embed this notice
But nostr handles that client side doesn't it? I need my client (amethyst) connected to at least one common relay with any given contact. Which implies a very large number of open connections to scale up (and an even bigger battery drain then now). On activitypub, the server handles that.
Or did I misunderstand?
-
Embed this notice
@c060b31fe2bbb0be4d393bc7c40a80848a25b8f0e0f382cb5b49c37bf7476cb4 @gvs They push more than they fetch, but they do a lot of both. And the answer in both cases it that "they just do." It loops them. Lot of servers? It loops them longer.
-
Embed this notice
How do AP servers deal with fetching so many events from other servers?
-
Embed this notice
They don't fetch, events are pushed and that takes a lot of resources
-
Embed this notice
How do the servers decide where to push each event?
-
Embed this notice
@c060b31fe2bbb0be4d393bc7c40a80848a25b8f0e0f382cb5b49c37bf7476cb4 @gvs When you post, the server publishes it to all the servers where people follow you.