I implemented something similar for another client and disagree
viciously ;) The proper timeout depends in the network, and asking the
user is a cop-out.
I am currently on vacation at a farm in Italy, very
nice and not just because of the free wlan, but the configuration of
the network here is not a user matter. You do not want to reconfigure
mutt to go on vacation. (Or perhaps you do, but let us pretend you
don't.)
A much better approach is to count seconds of connection
idleness, and after a few tens of seconds to mark the connection as
unreliable. When using an unreliable connection, the first command must
be a noop with a strict timeout.
Arnt
