> Spamasassin and retraining a false negative > > What I really want to do is something that TB doesn't > seem to support. i.e., I and lots of other people want > the ability to pipe a message to a program. I have > looked, lots of people have asked for years, and there > seems to be no addon to provide that capability. Ok. > Let's see if I can do something close. > > I have my own linux home server running sendmail. I have > dovecot running to be my imap server. I have spamassassin > and that's hooked up to sendmail via something called > spamass-milter which rejects spam before it even gets > accepted. The system works really well and I have not > seen a false positive > for about two years. [I also have clamav-milter running, > but that's not really relevant to this question.] > > The bottom line is that spam processing works very well, > but even so, I may > get a few false negatives per day. Those that do get > through need to be passed back to the > > sa-learn --spam > > command. If I don't do this, then there's no way for > spamassassin's bayesian database to learn. So the > question I have is this: > > Is there a way to shut off all spam filtering that > thunderbird might want to > be doing, but at the same time, allow me to click on a > message and then click on the "Junk" button and cause > that Junk button to run the sa-learn --spam command on > the message? I did see that there's a file that's a part > of my TB > 2.0.0.21 installation (under linux) called > /usr/lib/thunderbird-2.0.0.21/isp/SpamAssassin.sfd but > its content doesn't > give me any clue. > > And if this will generate any sympathy for me, what I'm > doing now is to save each message in a file and then run > my program using that new eml file as input. > > One last footnote: I do report all the spam that gets > through to both spamcop and knujon, but I do that from > the same script that also runs sa-learn. > > Anyone want to take a stab at this? >
I do it so, that a cron-job grabs all mail from my confirmed-spam folder on server, and learns it and reports to dcc, razor and spamcop. It currently does only the report part, but I think it learns it as --spam in the same process. In my IMAP client, I just drag all the spam to the confirmed-spam folder, and forget it. Basically I have 2 spam folders, Possible-spam and Confirmed-spam. All getting over 10 points from SA will go to Confirmed-spam, and all between 5-10 go to Possible-spam for a verify. Works for me.
