Well, couldn't help myself... I'm new to dspam, but: there's been reports of well over 99% accuracy in detecting spam with one instance of dspam alone. How would two dspam "layers" be able to achieve better? I'd say, there's no way.
And what's wrong with well over 99%? I'm happy, my users are happy. What more could you ask for? DSPAM rules. Just my 2 cents worth... /Lars Tom Allison wrote: > Under postfix the typical deployment for dspam is to set it up as a > content_filter. > > dspam has more than one way to score spam (eg: chi-square, markov) which > I've been told shouldn't be mixed (eg: CRM114 + chi-square). > > If (here's the question...) > > I were to set up a chi-square content_filter on port 10020 and send all > spam through that first with the option to quarantine or deliver to port > 10021 > > AND > > I were to set up a CRM114 markov content_filter on port 10022 and send all > spam through that deployment with the option to quarantine or deliver > to port 10023 > > Would dspam work or kind of self destruct... ? > > eg: > > postfix--> dspam/bayes --> postfix -->dspam/CRM114 -->postfix --> lmtp > > This is what I would like to do... > > Now, here's the reasons why I would do this twice: > > bayes is largely effective. > > bayes is much faster than CRM114. > > CRM114 is also effective but slightly different. > > It is my opinion that multiple different approaches is a good thing to > do. This is frequently seen with multiple spam scanning tools > implimented (bogofilter, spamassassin...) > If I could do this using dspam in both methodologies than I would have a > single point of quarantine and management. > > What are the chances that I could do this without doing data integrity > damage or suffering other inconsistencies in performance? >
