Re: SpamAssassin integrated with MailScanner, using per-user configuration in SQL or otherwise?

[rn...@xoxy.net]

On Wed, 2005-07-13 at 12:19 +0100, Martin Hepworth -
martinh@solid-state-logic.com wrote:
> rns.spamassassin.n.semba@xoxy.net wrote:
> > Hi,
> > 
> > I know this is not the ideal location to ask this, as it IS more a
> > MailScanner question, but shall ask in case anyone here has experience
> > with it.  
> > 
> > I'm researching integrating SpamAssassin into a MailScanner setup, and
> > from reading the documentation for MailScanner, I get the impression
> > that due to the way MailScanner calls the SpamAssassin Perl module,
> > MailScanner will perform all scanning using one single user.  I haven't
> > yet seen a way to specify to MailScanner to pass the necessary arguments
> > to cause SpamAssassin to use per-user configurations (either file-based
> > or SQL-based).  
> > 
> > Is this in fact possible?  
> > 
> > Appreciate any help.
> > 
> > Roshan
> > 
> Roshan
> 
> in a word 'no'

In that case, what would the pros and cons be of running SpamAssassin
through MailScanner, vs. running SpamAssassin outside of MailScanner, in
an exim environment.  

One argument I'd seen before is that with the spamc/spamd combination,
if spamd broke, spamc would hang indefinitely, but I see that the latest
versions of spamc don't seem to have this problem, with the default '-f'
safe-failover option.  They also claim that as a result of the
SpamAssassin.pm being called directly through MailScanner, that the
performance is far higher than that of using spamc/spamd or multiple
spamassassin invocations, but of course there aren't any benchmarks on
this.  

What other issues should one consider?  

Roshan