Not really - you still need the centralised signalling server for initial connection setup between peers. And this is needed for every new connection.

Yes. And?

It's like you'll also need DNS, which isn't decentralized either.

If someone wanted to write this as a WebRTC-based decentralised version there is nothing stopping them using something the like the public signalling servers that Mozilla uses for their Hello service.

Also, if some users wanted to run a signalling server themselves, that's ok too.

I get your comparison, however DNS is not actually that fundamental to P2P networks, all you need are IP addresses (at least not after the bootstrapping phase). Whereas this is making the entire connection setup dependent on these signalling servers.

You're right, there are public signalling servers and such. I just wish we weren't so restricted in setting up direct connections, such that you could do it through your bootstrapping peer in the network.

Hmm. I still don't get it. You mean build another project entirely, but base it on a web browser using WebRTC?

I think it's a terrible idea for many reasons. Cite FriendFeed, Diaspora, Identica etc.

But at the same time, if someone wants to try: yes, it needs to be browser based, and the lack of a server doesn't need to be a limiting factor anymore.