pull down to refresh
Yep, a "Mesh" implies the relays talk to each other. They Don't
I can still imagine that "routing" or finding a certain client might become difficult if the number of relays and clients is large.
Nah, it happens organically. If you're looking for someone even semi famous, people you know can tell you one of the relays they use, because they will be widely followed. Everyone else is someone you want to follow because you met them in person, where you can ask them to tell you one of the relays they use, at the same time as you ask them for their username.
I didn't realize when I wrote the above reply that Nip35 assists with user discovery by advising nip5 username servers to specify a relay for each user: https://github.com/nostr-protocol/nips/blob/master/35.md
That way someone can tell you "I'm so-and-so@example.com" and when you type that into your client, it can automatically look up what relays they use.
That has nothing to do with the protocol itself. It is not a mesh network lol
nvm, my bad. I thought you were replying to me
It’s easy to imagine new services being developed to sit between relays to deduplicate content or otherwise optimize, or (even more likely) relays that are also clients of other relays for the same purpose. Either way I think it’s fair to say that a partial mesh or inter relay routing is necessary for nostr to scale to 1B, but maybe 1B is not necessary. Could be that smaller communities like an enterprise whose relay is behind a VPN, or topics/community-based relays will be the final form
And even if it was ... Facebook and the internet and Tor and "6 degrees of kevin bacon" have proven that almost everything is 6 or 7 hops away in a mesh.
Nostr isn't a mesh-based network.
You can just pay a relay via LN. In the end it's also PoW.