Re: [Assp-user] Replacing the bayesian engine with DSPAM?

[Kevin <ass...@la...>]

bytehd wrote:
> I admit I have not looked at DSPAM since I first was hooked on ASSP in
> February after reading the
> HOWTOFORGE article on ASSP. My Clients could not handle the spam load at the
> client anymore.
> (not a good idea to accept it all, then classify it, aka DSPAM)
> 
> Im going to go back as re-eval DSPAM, but as I remember:
> DSPAM is a Bayesian filter only.
> In other words, all those Hello, DNS, BL, PB, and incredible powers of PCRE
> checks
> in ASSP simply dont exist in DSPAM.
> 
> DPAM lets ALL the spam through, then tries to figure out if it is spam with
> ONLY ONE METHOD.
> the CRM114 Classifier written by a Univ prof.
> Bayes is also the most CPU-intensive method.
> 
> This is dumb design.
> Fritz and John have it right: 
> do the easy, simple-to-execute (helo checks come to mind) checking
> then, after all the OBVIOUS things are done, run Bayes.
> 
> DSPAM does NONE OF THESE Checks.
> It works hard to achieve its score.
> ASSP works smart to achieve its score.
> 
> Plus DSPAM has one option: on or off.
> ASSP has this:
> http://www.magicvillage.de/~Fritz_Borgstedt/assp/ASSP-Admin-GUI/
> http://www.magicvillage.de/~Fritz_Borgstedt/assp/ASSP-Admin-GUI/ 
> 

It's been discussed before, many times.
There has been no evidence, so far, that the DSPAM Bayesian engine is 
any better than the one in ASSP.
It's more complicated I'm sure, but does it work better?

Not to mention that DSPAM has it's own set of requirements and doesn't 
support Windows AFAIK.  (that alone pretty much kills the discussion)

I'd be interested in seeing any comparison you may do. :)

Kevin