I've finally caught up with the mass of emails from the weekend... :-) So, personally, I would favor Carlo's approach. I like having perfect technologies, and I agree that they have social implications. But obviously we can't direct all our efforts in that direction. So why don't we think about a future-proof design, but apply it to the quick PHP hack Blaine&Matt want to see?
I think we can handle this by abstracting the transport layer like Ted said - and focus on the HTTP module first. If the abstraction is "abstract" enough, we can think about implementing our own PSYCish protocol later, but would still be able to have quick results in PHP. This would probably even play well with an Elgg fork, depending on how they did it. -- S.
