New Beta 4.58.6 released

Glenn Steen glenn.steen at gmail.com
Mon Jan 29 14:48:46 CET 2007


On 29/01/07, Gerard Seibert <gerard at seibercom.net> wrote:
> On Monday January 29, 2007 at 04:00:48 (AM) Glenn Steen wrote:
>
> > On 29/01/07, Mike Jakubik <mikej at rogers.com> wrote:
> > > Julian Field wrote:
> > > > There was 1 serious bug in 4.58.5 which I have now hopefully fixed.
> > > >
> > > > *Please* give this version a try as I don't intend changing anything
> > > > before the 1st Feb stable release unless you notify me of any problems
> > > > or bugs.
> > > >
> > > > Support for the 'p' records in Postfix 2.3/2.4 will have to wait for
> > > > the next release of MailScanner, personally I think it is a nasty
> > > > bodge. I wouldn't be at all surprised if Weitse changed his mind on
> > > > this topic, he certainly should do.
> > >
> > > Julian,
> > >
> > > Could you give a little more detail as to what 'p' records are, and how
> > > this will affect MS's compatibility with postfix?
> > >
> > > Thanks.
> > Mike,
> >
> > Download the postfix source and read the comment in
> > src/cleanup/cleanup_milter.c (starts on line 114 in the version 2.3.6
> > sources). This is so far the best explanation of what it is I've
> > found.
> >
> > If we manage to implement "support" for it correctly, it'll not change
> > the support/non-support status of postfix one whit (Jules still
> > supports it, Wietse will still see us as more evil than the horned
> > devil for perusing the queue files directly... Well, perhaps not that
> > evil, but close:-).
> >
> > We aim at making "milter edited queue files" -> "a (new) normal queue file".
> >
> > If you want my patches to look at beforehand, I can get them to you in
> > a blink:-). So far they've been tested by me and Nerijus Baliunas (who
> > is a brave soul and seems to be running them in production:-). They're
> > good for 2.3 with milters, not yet 2.4 with body editing milters.
>
> I posted the original post regarding 'p' records on the Postfix forum,
> and these are two of the responses that I received.
>
> //Quote//
>
> He is referring to his mis-use of the non-public, undocumented Postfix
> queue file format and plans to keep track of changes in this format
> rather than abandon its use.
>
> Mailscanner attempts to implement a fast-path for messages that are not
> modified by the content scanner. It cuts too many corners to achieve
> this goal.
>
>         Viktor.
>
> //End Quote//
I'm not entirely sure I agree with Viktor here:-). There's no "fast
path" reason to how things are done in MS, nor does it "cut corners".
On the contrary, all messages are equal in a batch, and if one can use
the queue files for all other MTAs, why shouldn't one with Postfix
too? Redisgn MS to fit PF... I think not. Wietse once stipulated a
list of things that a system like MailScanner would have to do to "get
away" with reading the queue files directly. Turns out that Jules
already did every single thing on that list. So, apart from normal
things like this (they change their format to support something new,
we change our SW to suit), there really is no basis for the animosity
between the two camps.

> //Quote//
>
> It means that some people don't understand ELEMENTARY SOFTWARE
> ENGINEERING practice.
>
> I spend huge efforts to maintain compatibility with software that
> depends on EXTERNAL Postfix interfaces, even when MAJOR changes
> are made to Postfix. If something breaks anyway, then I will do my
> best to provide a solution to make it work again.
>
> Postfix queue files are an INTERNAL interface.
>
> Software that depends on Postfix INTERNAL interfaces breaks the
> warranty. It is unsupported. It breaks even with MINOR Postfix
> changes, and I will not provide a solution when it breaks.
>
>         Wietse
>
> //End Quote//

This is actually a little BS and a little true: Wietse publish his
code under the GPL, and have to live with the consequences. But he
doesn't have to support _our_ misstakes, when we do them.
Example: OpenProtect make a Qmail port of MailScanner. We _never_
handle problems with that... We _always_ redirect such questions in
the general direction of those responsible for that code/port.

And that (without all the moaning) should be what the Postfix crowd
should do when the problem is caused by MailScanner. No more and no
less.

It actually has nothing to do with what Wietse thinks is "good
engineering practices".

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


More information about the MailScanner mailing list