Developers are always short of interesting things to work on and business guys are always short of developers. There's a lot of potential for the connection process to be improved.
A video demo of "hey look, click here and something happens over there!" is all well and good for impressing your mother, but not entirely useful looking forward. They've mentioned concerns with horizontal scaling and yet they don't consider the obvious option - Erlang.
In their presentation for Pivotal's Tech Talks series, they dismissed XMPP on the basis of not wanting to learn about it(!). Their dismissal of onesocialweb seems inappropriate on technical grounds. Also, HTTP is not the be-all and end-all. Not everything has to be shoehorned onto HTTP...
I've posted to the diaspora-dev list pointing to this comment, so hopefully one of the Diaspora guys will take a look at this soon.
I'll also follow up shortly with a blog post with more details about how this might be implemented atop a well defined set of standards and the benefits of doing so vs re-inventing the wheel.
I'm really disappointed to see that the team has decided that instead of learning about an existing standard (XMPP) which fits the purpose perfectly, they've decided to begin defining their own protocol, which will go down the same path but with less rigour and a whole lot of wasted time. Eventually it will have to be trashed in favour of something more robust.
If that's what they've spent the last three months working on, they've had some bad advice. I thought Pivotal were all about "just get the damn thing done" and so would have encouraged them not to waste their time.