There will also need to be a stringprep profile specified for mailbox names when that step is taken. rfc3501 got through before that requirement was solidified, but UTF-8 mailbox names will certainly need a profile for interoperability. Should I be able to put U+00BD ("vulgar fraction one half") in a mailbox name on a system that uses '/' as the mailbox delimiter?
Actually, that problem comes in with modified UTF-7 mailbox names.
You're right that RFC 3501 squeaked in past the wire on the stringprep requirement.
I think that, given where we are today, a mailbox name is best considered as a sequence of octets, ignoring the "small" problem that different mailbox names will have identical presentation forms. Until we have a stringprep to bail us out of these things, we're stuck with muddling along with what common sense tells us.
Unfortunately, common sense isn't all that common. I bet that if I said "I think that you should only use precomposed characters for now, and don't use combining characters", someone will vigorously object... :-(
-- Mark --
http://staff.washington.edu/mrc Science does not emerge from voting, party politics, or public debate. Si vis pacem, para bellum.