Pyzor, Razor, and DCC?

[Matt Kettler]

At 08:46 AM 10/17/2003, Matt Kehler wrote:
>Forgive my ignorance..I'm fairly new to this :)  Can someone explain (or
>is there a good FAQ) the pros/cons of using Razor vs Pyzor vs Dcc?  I've
>checked the websites for them all, and searched the MailScanner list
>archives....but they all look fairly similar in function, and I am
>assuming using more than one of them would be redudant. Does one of them
>work better with MailScanner/Spamassassin than the next?  Is it simply
>user preferance?  This would be in a system wide config; not personal use.

Ugo already responded, but I figure I'll add in some more.

In general the drawbacks and benefits of a message-hash type system are
similar to that of a RBL, as Ugo said. It's a network check, so you have
the drawback of it being "slow" compared to the text search rules. However,
hash systems also provide a good way of helping deal with new spams that
aren't yet in the ruleset.

Personaly, I find that bayes is more effective, but adding on razor does
give a very nice "extra edge" in picking up the last few percent of spam.

If you want a comparison of the various hash systems and their hit rates,
look at the STATISTICS.txt that comes with SA. This will give you good
statistics about spam%, nonspam% and spam/nonspam ratios for all the rules,
including razor, dcc and pyzor.