I've been thinking about this for a while too, and I agree that the peer-to-peer element is essential. So far, it appears that all the would-be open social media apps/protocols/whatever rely on you having your own server. That, frankly, is a deadly failing. Per Metcalfe's Law, social networks are only valuable when they are maximally inclusive, and the requirement to have your own server is something that will always exclude the majority of (non-technical) users. The need for easy non-technical access means that there are only two viable paths for any social media application. Either:

1.) Go to web page and use application, or 2.) Download app and use application

This is how Facebook / Twitter / etc. do it., and they're clearly doing it right. But there's no law which says that the application needs to keep its data on a third-party server. Why not have it run a Peer-to-peer service?

In fact, I think that Peer-to-peer is especially suitable to social networking, given that social networks are typically highly clustered (eg., people who are my friends are much more likely to be each others' friends as well). This means that if the every client application is able to redistribute copies of friends' status updates as well as its own status updates, then a sync between any given pair of clients is likely to allow them to bring a large subset of their friends' statuses up to date.

Of course if you're rebroadcasting your friends' status updates as a P2P host, then the question of spoofing/authentication becomes paramount. Obviously that would have to be solved by having all status updates transmitted in encrypted formats which can only be decrypted / authenticated with a users' public key.

Oh, and once you've got dual-key authentication embedded in widely-used social trust networks, you could probably solve a lot of other authentication problems while you're at it.

All of this would need to be completely transparent to the user, of course. If it's any more difficult to set up and use than standard Facebook / Twitter, it will fail. But I don't see any fundamental obstacles to making it that easy.

Unfortunately I don't have time to do anything about this!