Duncan Findlay wrote: > The only negative rules will be: bayesian rules, bondedsender and > habeas. Figuring how to autolearn ham (non-spam) is the only obstacle > we still need to figure out. This is fairly off topic, but the other day I tired of downloading all my spam to check it for false positives, and so I whipped up a script to produce an index of a mailbox, with author and subject lines and sorted by SA score, and cronned it so I'd be mailed a daily summary to peruse. Today I noticed those summaries were getting spamassassing scores in the 30 range. I ended up whitelisting myself, though that doesn't feel like a good idea -- now SA might mislearn spam subjects as ham, and any spammer who forges mail from me will probably get through. Aside from bypassing SA entirely for local mail, is there any better approach? -- see shy jo
Attachment:
pgpjbVHYBiIuN.pgp
Description: PGP signature