Funny to see everyone claiming that relays were a mistake and we should have just done P2P.
P2P is what the world had BEFORE the relays were invented. And it wasn't getting the job done.
Relays, that don't only relay, was a stroke of brilliance. They're why we're here.
Not really. True P2P is hard, for group or social interactions, super hard.
The interesting protocols are Holepunch (used by Keet) and Waku.
The problem with relays is that it increasingly seems that their operators are held responsible for the content.
I don't consider that to be a bug, since everyone can have their own relay.
P2P is great for communicating P2P. Not so great for openly publishing the new Internet with millions or eventually billions of people on all sorts of applications.
I think that there are privacy solutions arising, including P2P, client authorization, group encryption, etc., but I am convinced that adding these things onto the protocol is more valuable than skipping the relays and going straight to them as the basis of the protocol.
The relays had to come first. Everything else is a plug-in technology.
Encryption is the way. If we can prevent relays from knowing what's in a message, they can't censor it. Unfortunately they also can't index it, and only people with keys can read it. There are problems to be solved, but the benefits of relays are worth it.
If everyone has a local relay, it's essentially P2P, you just call the database where notes are stored a relay. Nostr is still not very good in this scenario, the network relies on a few relays, if they don't work, notes disappear, etc. nostr:npub1melv683fw6n2mvhl5h6dhqd8mqfv3wmxnz4qph83ua4dk4006ezsrt5c24 was doing some interesting research on this.
I think we can get inspiration from protocols such as BitTorrent where adding nodes helps the distribution.
An idea: an experimental relay that listens on localhost, distributes events through Holepunch based network P2P. Not only archives them locally, but distributes interesting events to other such relays. All clients work well, they just think they speak to a relay (well, they do), but the way the relays distribute notes is through a P2P network.
The difference is that it isn't P2P by design, it can just be used as if it were P2P, by running a local server. It's still client/server, but housed on one machine.
You could also read/write only to a local relay and broadcast from there, over Tor. Then you could get around how slow Tor is.
They are but what do you believe is the next step?
I think we're moving to different layers/types of relays, and adding P2P side-channels, that aren't part of the original NIP-01.
But there's no adequate replacement in sight, for the original relay concept. That's been proven, now. It's a new network architecture.
