First, you can force the use of self-signed certificates at both ends, or you can even use a "pre-shared key" instead of a certificate for authentication. The intended shared key should be known to all interested parties.
Secondly, SSL is an encryption mechanism, so it has nothing to do with the development of a communication model, that is, client / server vs p2p.
Finally, if both of your p2p nodes have an open or live IP address, and each of them knows about the other, you do not need any central server. The real problem here is that in practice most computers are behind firewalls using NAT and private ip. Thus, you need some kind of mechanism to go through NAT, determine the public ip and port in order to successfully reach the target computer.
Take a look at the next SO question and answer it. It mentions some of the methods used by regular p2p applications, such as skype (knowledge of the skype protocol is limited as it is proprietary), gtalk, and other XMPP-based chat applications.