Postfix and Mailscanner sitting in a tree k-iss-ing

[paddy]

On Thu, Dec 30, 2004 at 05:45:30PM +0000, Drew Marshall wrote:
> paddy wrote:
>
> >Chances are it doesn't sit well with the postfix design, and so Wietse
> >is not willing to support it.  Naturally not supporting it means telling
> >people you don't support it, which entails explaining ... you can see
> >where this is going.
> >
> >Add a dash of historical teething problems.
> >
> >I still eager to be surprised by a description of why it would be
> >_technically_ difficult to implement a solution that could earn the
> >Wietse gold seal of approval.
> >
> >
> Problem is that any such solution would involve extra overhead. For
> example you could use some form of SMTP/ LMTP interface, either the Perl
> module (Like AMAVIS) or even a stripped down version of Exim or Sendmail
> but that means running effectively two SMTP servers, more to go wrong
> and extra overhead. After all if you wanted to use Exim you would full
> stop...

Precisely what I'm guessing.  It may seem absurd to you or me, but that would
seem to be the postfix philosophy, unless I've misunderstood.  No doubt
there is some value in such a discipline that is explained in that
phliosophy (I genuinely don't know, but I cheekily think I can guess).

> You have to pipe to a program but MS is not built like that

granted.

> and you
> would have to change the whole processing idea (And indeed lose batch
> processing and therefore the benefits of bulk scanning speed)

Surely some smidge of a program drops the files on the disk in a format
MailScanner knows, and the reverse the other side ?

Does this sound like bloat, compared to raiding the queues on the disk ?
Well, it does to me, but it doesn't sound like a huge technical problem.
Whether anyone would want to write, maintain, support or even use such
a beast ... well that's another thing :)

> and the
> potential security of allowing the MTA to queue messages in a back log
> should it become over whelmed due to a virus break for example.

Granted if you try to change the mailscanner pull into a postfix push,
without any buffer in between.  But as you point out, that's absurd.

Could also be fun getting certain kinds of failure semantics.
(At one time I think I had my systems, dual sendmail, set up
so that if MailScanner ever failed to start, mail would continue to flow.
I don't presently recall how or even whether its still like this,
never needed it :)

> You see the problem. MS doesn't fit because it has a unique design which
> was not thought about when Postfix (And indeed any MTA) was thought about)

No offence to yourself or Julian, but I fail to see anything very novel or
unique about using files in this way, even if its a jolly good way.

That postfix eschews this method, well I suspect that's another story.

> >My guess would be that noone is motivated to do so, and why should they be?
> >
> >
> I don't think it is motivation but simply recognition that square pegs
> don't fit well in round holes, so you have to make a decision that you
> like the peg and live with it's poorer fit (In some people's eyes) or
> you change one element (Hole or peg).

Yes! Each party is pursuing a best solution, it is simply that
their opinions differ as to what is best.

Suppose that the only uses for an interface using a sanctioned API
are things only of interest to the postfix developers, and the option
for users to debug postfix problems with postfix team support.  Against
the cost of implementing, maintaining and supporting another interface to
postfix, it would come as no surprise to me that such an interface doesn't
exist.  Indeed, the fact of its non-existence may well be evidence that
it simply isn't necessary.

That doesn't make it a big technical problem.

Hopefully, with these improved results, what we hear about the word on the
postfix list will become more along the lines of

'We hear that it now works well, but if you use the unsupported API,
we do ask that you go _there_ first for support in the event that
you have any problems'

Which is, after all, I'm sure all that was ever really meant.

Hell after gassing so much I've half a mind to try to hack something together
to prove a point, but I'll try to hide behind the excuse htat I'd only
replicate something trivial already done long ago.

Mind you, wouldn't it be fun if what they said was:

        "Can you try that again with the 'prim' interface?"

:)

Regards,
Paddy
--
Perl 6 will give you the big knob. -- Larry Wall

------------------------ MailScanner list ------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk with the words:
'leave mailscanner' in the body of the email.
Before posting, read the MAQ (http://www.mailscanner.biz/maq/) and
the archives (http://www.jiscmail.ac.uk/lists/mailscanner.html).

Support MailScanner development - buy the book off the website!