Aswath Rao, Andy Abramson & James Seng all point to Richard Stastny who identifies two papers written by Prof. Henning Schulzrinne from Columbia University that each discuss P2P based IP telephony (one on Skype and the other on implementing SIP in P2P form).
I recommend reading the papers if you have the time. If you want the synopsis Aswath presents a succint review.
I've always found the argument for pure P2P IP communications to be rather unwielding. Agreed, much of it makes sense in theory, but in reality there are far too many implementation hurdles to overcome and legacy infrastructure to retrofit to. Unfortunately, the migration to IP communications is gradual (not instant) and must be flexible enough accomodate the least common denominator.
So, basically, the true vision of serverless P2P apps living at the extreme edge of the network is not practical in this world since the network will evolve faster than the majority of the possible endpoints. The functionality of IP communications is limited to it's weakest user, which today is average joe on the PSTN. Unfortunately, I think it may take a while for all 6 billion average joes to leave the PSTN. So, we'll just have to make do with hybrid until then. So it goes.