Postfix and duplicate mails

[Leland J. Steinke]

Keith Edmunds wrote:
> On Fri, 29 Aug 2003 13:54:12 +0200
> John Wilcock <john at TRADOC.FR> wrote:
>
>
>>I also see "skipped,
>>still being delivered" in the logs as described in thread on the
>>postfix list.
>
>
> <aol>me too</aol>
>
> The techniques described in the Postfix FILTER_README (by default at
> /usr/share/doc/postfix-2.0.11/FILTER_README.gz) would seem to be usable
> by MailScanner, and would have the advantage of using a support Postfix
> interface. Has anyone looked at doing this? I may try to if no one else
> has.

We did it for about 5-6 months before the "official" support for postfix made it
into MailScanner.  It required two perl programs: one to take the message from
postfix and put it into a format easily digestible by MailScanner (remember,
this is before MailScanner officially read postfix queue files directly); and
the other to reinsert it into postfix after MailScanner had completed its work.

I still have them laying around somewhere, if you are interested.  Never mind,
the message should be in the archives.  I submitted it on 4 March 2003 with a
subject of "Re: postfix compatability?".  I would trust what I wrote back then
rather than try to reconstruct it all now, since we are using MailScanner's
native postfix support.  We have been having minimal problems with 2.0.13 and MS
4.22-5.

There were no bugs in our code that caused us to go with the "official"
postfix/MS interface; we felt that the new interface would give better
performance than what we were doing previously.  There was some inflation of the
maillogs, since messages passed through postfix twice.  On the other hand, our
scripts generated two sendmail-esque queue files per message, so the overhead of
rewriting postfix queue files now borne by MailScanner was completely side-stepped.

Anyway, I have gone on long enough.  If you have any questions, either Lindsay
or I can answer them on Tuesday (Monday is a holiday in the USA)


Good luck,
Leland