It's been proposed in a few different forums by a number of different
people. Each version has had the problems that Tony Finch noted. I think
the most recent example of this idea was posted to the ietf(_at_)ietf(_dot_)org
list about 3 months ago. (No, I'm not going to go looking for it.) There
was a similar proposal at the FTC summit (patent pending, no less!) last
Bruce Lilly wrote:
A two-stage transfer whereby the message stays on the sender's host
until retrieved by the recipient (who is sent some sort of
notification that a message is pending) [such as one of Bernstein's
schemes] might balance the desires of sender and recipient.
However, that would be an entirely new architecture, has some
serious drawbacks for mobile users and other users w/o full-time
connectivity, and has not (to my knowledge) been written up even
in draft form.
I believe you are making unwarranted assumptions about a non-existent
design of an unspecified protocol.