@kate@social.treehouse.systems
Post #752344
2026-03-02 15:50 UTC
Replies (3)
-
@kate@social.treehouse.systems 2026-03-14 18:13
i figured it makes the most sense to just use OCM as the underlying mechanism, because nextcloud already implements and uses it. luckily in the case of federated sharing the system doesn't really work like a distributed system, because there is always one node that has the correct data (the node where the share was initiated from) and the others just follow it. the data getting out of sync is not really an issue, especially because security relevant things will always be delegated to the origin instance, thus not relying on trusting other nodes to apply the restrictions correctly.
-
@bootie_fringe@girlcock.club 2026-03-02 22:37
@kate everything XMPP that is usable across multiple clients. in particular, encryption (OMEMO), msg acks, stream management. I have had other chats drop texts or duplicate them – even after being in a train going through a tunnel XMPP managed to fix the chat state. It is important too to look at what did not work well – e.g. in-band encryption (OTR), having to update every server/client in lockstep, stuff matrix does that makes 1 implementation the “blessed” one or leads to performance issues.
-
@piegames@flausch.social 2026-03-03 00:20
@kate what kind of protocol and what kind of federation? Though generally, what you want is some 101 on distributed systems. (Matrix was right about saying a federated chatroom is nothing but a distributed database, even though they took the very wrong conclusions from that)