The IETF is the wrong place to go about design, so here goes.
It seems to me that S/MIME already has the functionality of a dropbox
scheme. What is lacking is what the usability folk call the 'affordance' of
being used in a dropbox fashion and well, usability.
Further, for such a scheme to become widely used it will be better if the
scheme is a protocol that can be supported by existing service providers as
an extension to their current offerings rather than a completely new scheme
that attempts to disintermediate them.
There are two separable forms of authentication required in such a scheme.
First we have the the access controls to the remote mailbox, second we have
the cryptographic controls on the data. My bits may b encrypted but that
does not mean that I want just anyone to view them. And even encrypted bits
typically leak some metadata.
* The recipient will need a public key against which senders can encrypt
their data.
* Senders will require a mechanism that assures them they are using the
correct key. If the parties know each other, a 'strong email address' would
be sufficient, that is a fingerprint of the public key combined with the
destination email address. If the parties do not know each other, a Trusted
Third Party (or parties) will be required to provide an introduction.
* The mailbox provider can map the key identifier to the public key if
necessary.