David Schweikert:
> On Thu, Oct 22, 2009 at 11:52:36 -0400, Victor Duchovni wrote:
> > Which version? All recommended patch sets applied?
>
> We have Solaris 10 (kernel patch 138888-03, dated January 2009).
>
> I did now a tcpdump during another such case and I found out that this
> is happening during the DATA phase: the client is slow and it takes 30
> minutes for it to transfer all the data. At the end, it manages to send
> a '.', after which, the "Service unavailable" is sent by Postfix.
>
> Shouldn't Postfix react after smtpd_timeout and not wait until the end
> of the data?
smtpd_timeout is no MESSAGE TRANSFER time limit.
smtpd_timeout is an INACTIVITY time limit.
If it were a message transfer time limit, then the message size
limit would be inversely proprortional to the bandwidth between
client and server, and that would be a mistake.
Yes, this means that clients can send messages at 1 byte per
$smtpd_timeout seconds. It is not a problem at this time, and when
there are reasons to expect that a session time limit is in order
then I'll do something about it.
Wietse
