I'm happy to hear that your product will try to avoid
confusing the user, but I'd feel much more comfortable knowing that
any compliant implementation is required to do as well.
This is not unreasonable; perhaps a policy document (similar to the Hosts
Requirements RFC) detailing PEM-related policy and UI concerns would be in
order (I'd certainly be happy to contribute to such an effort, in fact).
However, I still see this as distinct from the MIME/PEM representation
issues that I understand the current proposal to be addressing.
Also, I worry
about the home user market that you cited. I think these folks are
being set up to deal with a key management problem that will quickly
exceed their ability to understand what is going on. For example, how
confident are you that the home user will understand that an encrypted
but unsigned message do not really mean that the originator is
accurately identified by the "From" field in the header?
This, I think, depends completely on the UI, documentation, and training
they receive, not the underlying standards their software is based on.
Also, casual users tend (according to the feedback I've been receiving
so far) to only be interested in privacy in correspondence with a relatively
few number of people, with whom they are relatively content to exchange
keys manually (this is not to say that good key management isn't important,
just that the patterns of communication are different from the ones that
many of us take for granted).
But to take the example you cite, my own reply would be "if they get an
encrypted but not signed message, and the key isn't associated with a user
in their address book, I'll put up a message saying that the originator
could not be positively identified." On the other hand, designing good UIs
for Internet software is my job, so I may be biased concerning the
feasibility of trying to mandate good UIs in the standards documents.
I certainly don't think it belongs in the one currently on the table,
although as I mention above I think a document that concerns itself
explicitly with policy would be a very good thing to place in our input queue.
Amanda Walker
InterCon Systems Corporation