On 01/07/05 05:03 PM, William Stearns sat at the `puter and typed: > Good evening, Scott, > > On Fri, 7 Jan 2005, Scott Wertz wrote: > > > I think this is an easy question, but I haven't been able to find an > > answer. If I'm using spamassassin 3, invoking it via procmail as just > > 'spamassassin' and testing for the result, and I trust that any message > > carrying a URL that's listed on surbl.org or spamhaus.org is 100% spam, > > what file(s) would I edit and how? > > > > In other words, I've never seen a false positive on either of those BLs, > > but I'm seeing spam that meets those tests and is still weighted less > > than 5. I want to change that. > > As Michele correctly pointed out, you're certainly welcome to > drive up the scores quite a bit so that emails with an surbl-listed domain > are much more likely to cross 5.0. > However, even though Jeff Chan will likely shoot me for saying it > ;-), surbl's can and occasionally do have false positives. Let's use > Gevalia coffee as an example. I'll blacklist their domain because they > regularly send me UBE. However, Gevalia has legitimate customers; for > those individuals, email from that domain is _not_ UBE, it's solicited > mail. (Just for reference, we removed gevalia.com because there were > legitimate uses for it...) > I personally have trust in the surbl's, so I have no problem > recommending that people increase the score if they want. Might I humbly > recommend increasing the surbl score to something between 2 and 5, so that > if surbl screws up for your particular mail flow the other rules have a > chance of reining it in?
Excellent suggestion. Shoulda made it myself in my other posts. It's been mentioned a number of times that scoring a test at 100 is almost always a bad idea. Boosting the scores to allow them to swing a bigger bat - although one that can potentially be overridden by very low bayes scores - is usually ok. Lou -- Louis LeBlanc [EMAIL PROTECTED] Fully Funded Hobbyist, KeySlapper Extrordinaire :) http://www.keyslapper.org ԿԬ Information Processing: What you call data processing when people are so disgusted with it they won't let it be discussed in their presence.
