Why not adopt the goal of deprecating all ciphersuites but one? Let's have a debate about the best ciphersuite and settle on a single one. That would decrease the attack surface and would allow the crypto community to focus on getting the implementation bug free. It would also give attackers a tighter focus, so we'd better pick the right ciphersuite! To Brian Smith's list of criteria, I would add implementation simplicity as a very important requirement.
For the past 5 years or so, we've managed to take good advantage of the fact that we had a diversity of ciphersuites, even though each of the pre-1.1 ciphersuites had problems. Trying to lock down a single known good ciphersuite would be drawing exactly the wrong lesson from recent history.
If the latest Greenwald/Schneier revelations are true, then the past 5 years of TLS have underwhelming to say the least, so I don't see what advantage has been obtained from diversity.
Going forward, how are site admins to know what ciphersuite to use? This needs to be made very clear. And how are web users to know what the green lock in their browser means, if it can mean anything at all? The current state of affairs is incomprehensible to all but a very few.
There needs to be a big public debate about what ciphersuite is the best. And there should be very careful scrutiny of the implementation. Choosing ciphersuite(s) is a political as well as a technical problem -- the goals should be as clear and simple as possible.
