Storage: service provider vs user

I once asked someone from an ISP why they only support POP access and 
not IMAP. His answer: "we want users to download their mail". Users on 
the other hand like storing their mail on the server. I think a new 
protocol should include semantics to handle these conflicting goals, so 
that the user agent and the server can negotiate who stores what. For 
instance, the server may be prepared to store mail for a limited time 
or a maximum amount. Even if they don't get to store mail on the server 
forever, it's still very valuable for users to be able to keep the mail 
on the server a little while after looking at it for the first time. 
For instance, the user may be connecting over a slow link and not want 
to download everything just yet, or the user uses several computers to 
read mail and wants to make sure each of those has a copy before the 
message is removed from the server. A way to track how many clients are 
supposed to get copies of a message and how many have already would be 
useful here.

In Fidonet and I think also UUCP messages are bundled in an archive so 
that someone connects it's easy to download all new messages. Do we 
want something like this in the new system?

And would it be useful to create two levels of mail access: the first 
level means someone gets to download the mail but it's encrypted, and 
the second level means the ability to decrypt it. This way you could 
ask a friend to download your mail for you, but they can't read it. 
Something like this could be useful for intermittently connected 
clients.