Today, for the first time, spamassassin's bayesian filter decided its own cron job output wasn't spam.
I've been running 0 3 * * * /home/darxus/sausr/local/bin/sa-learn --rebuild since 8/26, and every day it was ranked "spam probability is 90 to 99%". I was about to give up on spamassassin and bayesian filtering and switch to purely crm114 (yes, I realize it's very related). Yesterday I specifically filtered down to each of these cron output emails, and ran them all back through sa-learn as ham again. Turned out I missed some. And, as I said, today it got much better: BAYES_10. So I'm guessing I must've accidentally flagged one of them as spam, and that screwed up the data enough that it stayed confused until I reprocessed them all as ham. $ ls -l bayes_msgcount -rw------- 1 darxus darxus 4049 Sep 16 14:58 bayes_msgcount (I've been processing emails... all ham, and missed spam) <just barely avoided asking a question from the faq: #5.2> -- "Every normal man must be tempted at times to spit upon his hands, hoist the black flag, and begin slitting throats." - Henry Louis Mencken (1880-1956) http://www.ChaosReigns.com ------------------------------------------------------- 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
