Peace proposal: postfix+sendmail+mailscanner

[Glenn Steen]

On 10/03/07, am.lists <am.lists at gmail.com> wrote:
> On 3/9/07, Hugo van der Kooij <hvdkooij at vanderkooij.org> wrote:
> > Right,
> >
> > In this holy war of MTA's may I suggest a peace proposal? At least humor
> > me and read it in full before you start firing again.
> >
> > Just a few assumptions:
> >  - MailScanner is designed to work on mailqueues. (batch wise)
> >  - sendmail's way of working with mailqueues does fit MailScanner.
> >  - postfix prefers SMTP conversations between postfix and 3rd party
> >    components.
> >
> > As some prefer to maintain postfix configurations instead of sendmail
> > configurations there is a gap left between the MailScanner and postfix
> > ideologies.
> >
> > How about using a stripped down sendmail config to bridge that gap? That
> > config could be generic as far as I can figure it out.
> >
> > The flow would then be something like:
> >
> >        Outside world ==SMTP==> postfix:25
> >        postfix ==SMTP==> sendmail:10225
> >        sendmail ==QUEUE==> MailScanner
> >        ...... MailScanner (doing a lot of tricks)
> >        MailScanner ==QUEUE==> sendmail
> >        sendmail ==SMTP==> postfix:10025
> >        postfix delivery as usual
> >
> > Basically sendmail would be a stripped down solution to translate from
> > SMTP to inbound queue and outbound queue to SMTP again.
> >
> > >From the postfix perspective sendmail is just another filter pretty much
> > like one would do for amavisd, .....
> >
> > The trick is to build a sendmail config that would do the trick and does
> > not require changes for multidomain handling. The SmartHost option would
> > be sufficient to handle the hand of back to postfix. (Even I can write
> > that part. ;-)
> >
> > Hugo.
> >
>
>
> Pardon my ignorance here, but are we attempting to get around Wietse's
> ALL CAPS statement that MailScanner uses unsupported methods to
> manipulate Postfix?

Hi Hugo & Angelo,

You just had to spoil the friday night fun we were having by turning
this serious, now didn't you:-).
Read on, I'll be very serious indeed...

> If that's our goal to solve, why don't we simply have Jules and Wietse
> have a conference together and discuss the access methods. Perhaps
> Jules is doing this 'unsupported' access in a completely harmless way,
> such that if Wietse understood it more fully (I'm not suggesting that
> he doesn't...) then perhaps Wietse would be more forgiving and amend
> his statement that "as of such and such version, MailScanner uses
> supported methods..." or something.

This has been proposed a few times, but neither Jules nor Mr Venema
has shown any inclination to take up any form of negotiations.
If you search the archives very thotoughly you'll find a few threads
were we've discussed this in the past (Dahwal Doshy has been a brave
soul and tried to get something out of Viktor and Wietse a while
back)... In one you'll see that Wietse at one time let slip a "list of
demands" that MailScanner would have to fulfill to be termed
"marginally supported"(:-) ... As it turned out, Jules had implemented
his Postfix support exactly like Wietse stipulated it had to be (HOLD
feature style, mind you). Central in these "demands" were that one
couldn't copy the postfix queue file around, but rather had to
deconstruct the old one and reconstruct a completely new queue file...
Which is exactly what we do.
Now, last I looked the code for postfix is free to look at, so ... the
moaning about SW engineering practises and safe uses of published APIs
becomes less relevant or beleivable.
But, as it stands, Wietse will never endorse MailScanner in any way
shape or form, and this might not have that much to do with anything
other than him not liking things NIH, for all I know.

> I think having another layer for transport is not necessarily a good
> thing. It's one more thing to troubleshoot when things get ... well,
> you know how they can get.

Indeed.
When one looks at list archives, particularily the postfix users one,
one can stumble on some very old suggestions on how to overcome this
"unsupported usages" viewpoint... Like using a script (some FILTER
thing, don't remember all the details) to capture the messages to
another type of queue file (non-postfix, with envelope data preserved)
for MS to process. This approach never got much support from any camp.

Anyway, the "politics" page in the wiki
(http://wiki.mailscanner.info/doku.php?id=documentation:configuration:mta:postfix:politics)
 has become a bit old, but it still contain some of the historical
debate/ciewpoints.
It accurately pinpoint one central thing: This indeed is politics, not
technology.

> Just my $0.04 worth (that's ~£0.02 for those on the other side of the pond)
>
> Angelo

Cheers
-- 
-- Glenn
email: glenn < dot > steen < at > gmail < dot > com
work: glenn < dot > steen < at > ap1 < dot > se