Several of us have recently thought that a PGP finger print makes a fine name, but this is in the context of some implicit assumptions. Here are some of those: First I want to propose a requirement on a name space that is weaker than “collision free”:
The entity wanting to gain a name in the name space should have a feasible way to acquire a name that is and will remain unique even after he disseminates the name among untrusted agents.
Tyler raised the issue that knowledge of a name should not convey authority to send to the named person. This connection results from one of the above assumptions that there is a public registry of public keys. That need not be. This is notion of keeping public keys secret. If we decide that the common presumption is indeed good (I think it is) that the public key of a RSA key pair is indeed public, then I can identify to you Joe’s identity by my sending to you the hash of the hash of Joe’s public key.

See this. Some new stuff