On Tue, 2003-02-25 at 17:45, Mark Crispin wrote: > However, it does seem that using the IDLE extension will provoke a NAT > timeout, so a client that implements IDLE should have a means to disable > IDLE in favor of polling.
Why not just restart the IDLE more often? Is sending DONE + IDLE once every 3 minutes worse than sending NOOP every 3 minutes? Except for the bandwidth usage being a bit larger.
