4 ms·

LOL that was a Procmail flashback from twenty years ago! What happened to Bayesian detection and AI?

• dpifke 4y ago
  They work, too. But sometimes simpler solutions work better. SpamAssassin assigns a score (positive = more spammy, negative = more hammy) to each rule. So the above-quoted rules work in conjunction with other rules, the Bayes filter, IP/ASN reputation data, etc. DWP_419 has a really high score in my setup right now because the false positive rate is currently zero, but it could theoretically be canceled out by other signals.