Jeremy Kister <[EMAIL PROTECTED]> writes: > While coding, I had a thought... Would it be a good idea to > Auto-White list/Auto-Black list Mail Servers themselves? While most > of us don't have access to MAPS, it might be a good idea.
It's a great idea. It's been mentioned a few times (I mentioned about the same idea in bug 1084 as a better solution for another feature request). Anyway, I think the right implementation would based on the current whitelist code so there's scaling based on SA score and other factors. I'm now leaning towards using the From: address complete hostname plus the first half bits of the IP address (basically, just like the current AWL minus the "username@" part). I would probably use less magnitude than the current system. Since this will also overlap with the current AWL system, you could lower the magnitude of both, something like: (full address + first half of IP) AWL = adjustment * (2/3) (hostname + first half of IP) AWL = adjustment * (1/3) So, 1/3 of adjustment is host-based and 2/3 of adjustment is full address based. Or, you could always store and increment both entries, but only use the hostname one if the full address one has seen less than 5 messages *and* the hostname one has more entries. Otherwise, only use the full address one. The RIGHT way is, of course, to experimentally test both methods on real email and see what methodology, constants, etc. yields the best possible FN:FP ratio. Is it possible to use mass-check with a mass-check AWL (that won't interfere with the one in $HOME)? > Thoughts? Want to open a bug report for this and submit a patch? ;-) Dan ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
