On Sun, Oct 05, 2003 at 10:36:51AM +0200, Hendrik wrote: > First Question: Why does SA prints out "autolean=no" when the > /etc/mail/spamassassin/local.cf says: > > # Enable Bayes auto-learning > auto_learn 1 > > ???????????
It does not do auto-learning on every mail. Imagine if it learned a mail that scored 5.1 as spam when the mail was false positive. Or if mail scored 4.9 and it was spam. Only tokens from mails that are clearly spam or good mail are automaticly inserted into bayes database. There's score limits (-3 and +15 or something) but I think there's other mechanisms to help minimize auto-learn mistakes other than just scoring-based. > After copy the mail to my linux box I started spamc: > > spamc -r < Unbenannt.msg > 4.3 HEAD_ILLEGAL_CHARS Header contains too many raw illegal .msg means it's saved to a file from Outlook, right? SpamAssassin can't parse those. Copy headers from the mail's options in Outlook and body (if html, copy the source) to a new file (so it'll be just ascii instead of binary garbage what the .msg probably has) and try again. -- (Mr.) Hannu Liljemark | Appelsiini Finland Oy | http://appelsiini.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