Stan Hoeppner:
> > Agreed. Postscreen's main goal is to reduce mail server load, so
> > that you can postpone that forklift upgrade.
> >
> > Postscreen also stops a few percent of spambots that popular DNSBLs
> > miss, but at this time, that is only a minor benefit.
>
> I would think the proper metric for evaluating the success of Postscreen
> deployment should be something like
> mx_#smtpds_per_connect_per_day_week_month vs the period before deploying
> Postscreen; load average before and after Postscreen, Postfix memory
> consumption, etc. It would include no spam catch/miss/false
> positive/negative data as the difference between before/after would
> likely be within statistical margin of error.
>
> Has anyone compiled such data? If so and I missed it, apologies for
> having my head in the sand.
On my tiny site, spam volume turns out to be more variable than I
expected, so comparing before/after differences is not so simple.
One way to cancel the variability is to run equal-preference MXes
with different configurations. Another way is to randomly switch
configurations several times a day.
I expect, though, that the exact numbers would be site-specific.
Wietse
