SpamAssassin/DCC integration

Craig R Hughes <craig@hughes-family.org> writes:

> Matt Armstrong wrote:
>
> MA> I'd say if performance is a problem, then SpamAssassin should be
> MA> written in a language faster than Perl.  The fork/exec + pipe of
> MA> the message into dccproc is not going to be a huge bottleneck
> MA> for SpamAssassin users.
>
> At the risk of repeating myself....
>
> It's not so much the CPU performance, it's the RAM usage level
> performance issues related to having to copy every message between
> processes.  The CPU issues are also of concern, but as you point
> out, there's plenty of optimization to be done in the way SA
> currently works, and there is work in that direction being done.

I'm simply arguing the overhead is not a bottleneck, while you're
arguing that it is.  Neither of us are actually backing up our claims.
;-)  Only benchmarks can provide an objective answer.  Have any been
done?

-- 
matt