On Fri, Nov 09, 2001 at 09:23:05AM -0500, Joseph Kezar wrote:
> another copy? Is that so? If the message is marked as "D" for that user
> it should skip them though. So I thought.
This is basically correct. And your analysis is wrong.
The SMTP protocol is lock-step (except by use of the PIPELINING extension),
and for the most part, knowledge that a timeout happened will cause you to
know that a message was either delivered or it was not. However, there is
one state for which this is not true.
>>> DATA
<<< 354 Send data followed by <CRLF>.<CRLF>
>>> From: ....
>>> To: ...
>>> Subject: ...
>>> Other-headers: ....
>>>
>>> body text
>>> .
At this point, the receiving end must attempt to spool or deliver the
message. It will then return
<<< 250 Accepted id=...
If the sender times this out, the message may well have been delivered, but
the receiver never gets a chance to tell it, so the sender *has* to assume
that it hasn't.
I can't remember offhand which variable this is in Exim, but I find the
default timeouts to be sensible and not to cause this to happen in general.
MBM
--
Matthew Byng-Maddick <mbm@???> http://colondot.net/
