Comparisn with DSPAM

[ram <ra...@netcore.co.in>]

Anyone using DSPAM ? 
I have never been a fan of bayesian methods. But recently I have seen
references to DSPAM. 

Can I use DSPAM along with spamassassin. I use spamassassin as a module
in MailScanner. What are the pros & cons of using both.

Thanks
Ram

Re: Comparisn with DSPAM

[Matt Kettler <mk...@verizon.net>]

ram wrote:
> Anyone using DSPAM ? 
> I have never been a fan of bayesian methods. But recently I have seen
> references to DSPAM. 
>
> Can I use DSPAM along with spamassassin. I use spamassassin as a module
> in MailScanner. What are the pros & cons of using both.
>   
I don't think that part makes much sense if you're thinking of it from a
perspective of having SA use DSPAM the way it uses razor or dcc... SA
already has a bayes subsystem of its own, so why would it make use of DSPAM?

Sure, you could use SA/MailScanner at the MTA level, then do a second,
independent round of spam scanning with DSPAM at the MDA level (which is
where DSPAM wants to be). But I don't know if this would be particularly
"worth it". You'd be adding a lot of overhead, and you'd now have two
bayes databases that need manual feeding to work optimally.

Personally, I view dspam (or bogofilter for that matter) as a tool for
those who either can't run SA due to the overhead, or just want a
simple, straightforward bayes-only scanner. SA's multi-rule based
approach has advantages (resistant to various forms of poison tactics),
but also has disadvantages (complexity, overhead) when compared to
single-rule based systems.

If you just want to "add bayes to the mix", I'd suggest starting off
with SA's built-in bayes. After all, you've already got SA installed. If
you're not happy with the results of SA's bayes, or need to shed some
overhead, by all means, go look at other tools.