Comparison of different DNSBL's

[cleverhandle]

The recent thread on IP blacklists reminded me to ask this - for those of you with experience using different DNSBL's for spam prevention, how would you say they compare? I recently set up my server to use spamhaus's sbl-xbl, which seems fine. But it's only been a couple of weeks and I don't get much traffic, so I don't think my experience counts for a whole lot. I tried Googling around for some comparisons when I was setting up my BL, but didn't find much apart from isolated complaints. But I'm guessing that some are more conservative than others, or that some target certain sources more effectively. Suggestions?

 

[Need4Speed]

here are the restrictions and DNSBL's I use, along with Spamassassin and Net:NS module.

from my postfix main.cf file:
smtpd_recipient_restrictions = permit_mynetworks, reject_invalid_hostname, reject_non_fqdn_hostname, reject_non_fqdn_sender, reject_non_fqdn_recipient, reject_unknown_sender_domain, reject_unknown_recipient_domain, reject_unauth_destination, reject_rbl_client cbl.abuseat.org, reject_rbl_client blackholes.easynet.nl, reject_rbl_client sbl.spamhaus.org, reject_rbl_client opm.blitzed.org, reject_rbl_client relays.ordb.org, reject_rbl_client dul.dnsbl.sorbs.net

I've trained the bayesian filter with tons of emails (spam and ham) and I would say that I'm up to 98-99% effective filtering.

here is a quick RRD graph:
http://www.cyberfrogs.net/rrd/graphs/

the RED are rejected at the MTA level
the ORANGE are rejected by Spamassassin

You could conclude from the graphs that the MTA rejection via the DNSBL's isnt all that good when compared to SA.

Here is a list of the tests and DNSBL's that SA uses:
http://spamassassin.apache.org/tests_3_0_x.html

You can see that something drastic changed in my filter around the middle of november...thats when I added the the Net:NS module.