Russ,
Thanks for the comment. There's more work to do before the spec is
completely ready to become an RFC, so we'll add words to that effect.
It's certainly the intent that when someone wants to sign and encrypt
a message, they should sign it first and then encrypt it.
That said, it's not completely obvious to me that signing an encrypted
object is *always* useless. I think there are some plausible, albeit
infrequent, scenarios in which I might want to forward something that
happens to be encrypted. This is particularly true if the encrypted
object was encrypted by someone else and I want to forward it with my
signature protection.
Steve