Re: Misc minor POP3 issues

[Michael D'Errico <[e-mail]>]

John Gardiner Myers <[e-mail]> writes:
>[e-mail] (Michael D'Errico) writes:
>
>> This is an issue related to the broken TCP connection argument we've
>> been having.  I implement a timeout which defaults to 10 min. and is
>> configurable.  If the server times out, it sends a "-ERR" message and
>> closes the connection.  The UPDATE state is not entered, and no messages
>> are deleted.  Does this seem like a reasonable thing to do?
>
>An autologout timer should certainly do the same thing with respect to
>the UPDATE state as a broken TCP connection.
>
>A potential problem: When the server times out, if it sends an "-ERR"
>message, a client may erroneously deduce that the error was caused by
>the next command it tries.

I agree, and I will be removing the "-ERR" response.  This was motivated
by the "421 closing connection" reply in SMTP, but doesn't fit with POP
since all errors are treated the same.

>If we permit an autologout timer, we should specify a minimum value
>for it so clients can know how often to send NOOP keepalives.  For
>comparison, the shortest permitted autologout timer in IMAP is 30
>minutes.
>
>I have seen a situation where a PC crashed while holding an open POP3
>connection.  The connection thus never got closed and the maildrop was
>left locked.  An autologout timer would solve this problem.

I think it's clear that an autologout timer is necessary and should be
included in the spec.  Specifying a minimum of 30 minutes seems fine
with me, and this would be consistent with IMAP.  Consistency is good.

Michael D'Errico
[e-mail]