Hi, I guess everybody knows it ... users hate spam. We get blaimed for it. Bayes could correct for some of the spam they get but they dont know how to use the webinterface or are too lazy to do it.
It has been posted before on the MailScanner and probably also the spamassin list. Since exchange mangles with the mail when using forward the mail can't directly being used to train the bayes. With spam as well as ham. I'm a situration where I'm allowed to store all mail going though the mail gateway. So I'm thinking about the following: * Users can attach mail and send then to a address like: [email protected] and [email protected] * A deamon looks for mail there. mbox, Maildir, IMAP, POP3 what ever. * Looks for attachments and tried to find the MailScanner-ID: 5B3A0192865.AB19B which can then be looked up in the database. * Sends a rabbitmq message with the needed information. So what are the pros and cons here. Pros * I dont have to manually train the bayes any more * I actually think users can figure out how to forward messages as attachments to a email addres. * Less spam, since bayes is now bayes on more feedback from users. Cons * They might forward to a wrong address, ham to spam and vice versa, but they should just forward again to "forget" The biggest problem here is that I'm no expert in Python, but that can ofcause be learned for this little task or there could actually be used any languages since they only thing I need to interact with is rabbitmq. Andrew ... I think you are the best person to answer if this is doable or I have missed something here ? Ofcause, if I create it, it will be made available as a crontrib to Baruwa or as a standalone package on github/bitbucket. mvh Mikael Syska http://ifyoudo.net _______________________________________________ Keep Baruwa FREE - http://pledgie.com/campaigns/12056
