Its still somewhat slow with loading existing chats but Matrix has serious potential. It is already doing much better than the fediverse in terms of discovering niche communities.
Matrix 2.0 will put an end to that trait plaguing the protocol's viability. With the new "sliding" sync, clients will be able to fetch stuff much more efficiently (and more akin to Discord's API, for example) than before. Syncing an freshly logged-in account takes seconds instead of minutes and syncing messages when returning to the app is instant. They presented this at FOSDEM this year: https://www.youtube.com/watch?v=eUPJ9zFV5IE
Some rooms have so many user and so much chat history, that my own puny homeserver struggles to join them. I've tried all sorts of tweaks - from tweaking the PostgreSQL service, to using things like noatime etc. on the ext4 filesystem. This has helped immensely, but some rooms, like the Python room on matrix.org, brings my homeserver to its knees :)
Perhaps in the future the Synapse devs will improve the code so that mammoth rooms don't bring more resource-starved homeservers down. Alternatively I could throw more resources at the homeserver, but for my use-case, the 6GB ram and 6 cores I assign to the VM running the Synapse instance and the PostgreSQL service - and the IRC bridge heheh - is the bare-minimum I can get away with.
On saying all that, I like Matrix more than I like IRC, nowadays, and more folks should IMO get on Matrix.
(p.s. custom emoji's would be lovely on Matrix ;) )
conduit runs on pi zero -- though you can't directly migrate so that is of limited use to you. It would probably fail to join matrixhq or python room but, this is due to obvious resource constraints. Any reasonable horsepower and you won't have an issue.
The slow loading and tracking of who has read what are the two biggest gripes I have about the protocol.
If I'm joining a huge channel with thousands of people I _really_ don't care whether astroboy8756 has read my message or not. I don't want that traffic to enter my client at all.
