MCP Efficiency?

[Glenn Steen]

On 18/08/06, Glenn Steen <glenn.steen at gmail.com> wrote:
> On 18/08/06, Julian Field <MailScanner at ecs.soton.ac.uk> wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> >
> > Sorry, just checked the code. I am doing that already.
> >
> > The cause of the problem is that SpamAssassin does not appear to support
> > the way I am trying to use it. I want 2 completely separate instances of
> > SpamAssassin. One has all the normal SA rules as expected. The other one
> > has no rules or dns checks or Razor or anything at all, it *only* has
> > the few rules specified for MCP checking.
> >
> > I can't make it do this, while still keeping all the rules compiled in
> > both instances and every setup done and cached. The only thing I can
> > make it do to run the way I want, is to tell it not to pre-compile all
> > the rules. As a result it has to do a huge load of SA compilation for
> > every message.
> >
> > If Matt Kettler is around, maybe he could offer me some advice. I have
> > tried asking on the SA list several times, and they don't understand why
> > I would want my 2nd instance at all, so I never got any helpful answers.
> >
> This is exactly what I suspected from the few trials I did recently
> (was thinking of starting to use MCP). And that is where I got the
> idea to set dns as unavailable. Helps some, but not all.
>
> The problems lie solidly in the built-in defaults that you cannot override.
> What I think you have to do is to maintain a "cleaned" SA environment,
> and see to it that MCP uses it by way of a chroot thingie. Really
> icky:-(.
>
Replying to myself after actually reading your code and the code of
the spamassassin command, as well as engaging brain just a tad (I
might be fooling myself here:-)...
Correct me if I'm wrong, but whatr you do in MCP.pm should be
equivalent to calling spamassassin like:
# spamassassin -D --lint -C /etc/MailScanner/mcp
--siteconfigpath=/etc/MailScanner/mcp -p
/etc/MailScanner/mcp/mcp.spam.assassin.prefs.conf 2>&1 | less -e

.... or something very similar.
That shouldn't get the effect you cite, nor the effect I (wrongly)
observed. It should work IMO (so no need for extreme measures:-).
Or am I totally missing something here?

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