> (ie, fetching keys from keyservers, encrypting automatically, etc)
No no no no no! The keyservers have no authentication for key addition. Anybody can put up a key for any email address and effectively wedge themselves man in the middle.
Also, the paranoid in me (and probably more significantly, the keyboard-activist-in-the-safety-of-my-parents-basement) suggests that it might be wise to access keyservers over TOR.
If _I_ were involved with PRISM, the pipe running to pgp.mit.edu would be one of the most monitored connections around. "Hmmm, someone just searched for a PGP key for FedRegister - lets see what else that IP address has searched for, and what's in all the gmail inboxes that have ever been accessed using that IP address…"
Yes, I redact my above comment to remove that example.
However, I still stand by my statement that PGP UX is a client issue, not an inherent protocol issue (ie, it's fixable without abandoning PGP entirely).
No no no no no! The keyservers have no authentication for key addition. Anybody can put up a key for any email address and effectively wedge themselves man in the middle.