Henrik K wrote:
On Sun, Oct 11, 2009 at 01:10:17PM -0400, Adam Katz wrote:
  
Here are the default scores for the DNSWLs I know of:

RCVD_IN_DNSWL_LOW 0 -1 0 -1
RCVD_IN_DNSWL_MED 0 -4 0 -4
RCVD_IN_DNSWL_HI 0 -8 0 -8
RCVD_IN_HOSTKARMA_W -5 # (nondefault rule, Marc's suggested score)
    

You have to remember that Marc is automated and as such can never be as
fully trusted as DNSWL which has a completely different listing criteria. I
have to admit W is getting better though, nearing DNSWL_LOW rates. I'd like
to see many levels of W though and not just a single lump.

  
RCVD_IN_BSP_TRUSTED 0 -4.3 0 -4.3
RCVD_IN_IADB_DOPTIN_GT50 0
RCVD_IN_IADB_ML_DOPTIN 0 -6 0 -6
RCVD_IN_IADB_VOUCHED 0 -2.2 0 -2.2
RCVD_IN_SSC_TRUSTED_COI 0 -3.7 0 -3.7
    

These practically don't exist, if you look at ruleqa.

  
For what it's worth there are really only 3 serious white lists on the planet. I'm surprised no one is
testing the emailreg list. There are dozens of black lists. Doing white lists is actually easier than doing
black lists because there are thousands of servers out there that send nothing but good email. That have
good FcRDNS, they are static, and unlike the black lists IPs they aren't trying to be evasive. It's low
hanging fruit. On my servers if you are white listed your message just sails through the system.

One of my beefs with the spam filtering community is that there is too much focus on detecting spam and not
enough focus on detecting non-spam. We need more white lists and we need more white rules. A lot of what
I'm doing is because no one else is doing it. I'd love it if other people would get into the white list
business and do a better job than me. I'm really good at coming up with new and original ideas but others
are usually better at implementing them.

I'd love to have sources of IPs that send nothing but good email. It would be trivial to set up a system to
detect that and to collate results for several trusted reporters. If I have some people who were interested
we could set something up and a lot of good email could sail through the system with better accuracy.



Reply via email to