Robert Bartlett wrote: > From: Bowie Bailey [mailto:[EMAIL PROTECTED] > > Robert Bartlett wrote: > > > If someone forwards me email as a regular forward and not as an > > > attachment, could I still feed it to bayes as the user who forwarded > > > it to me and still be effective? Or will that cause problems because > > > the headers show as the person forwarding the email and not the > > > original headers? > > > > > > What I mean about username is the username switch in the sa-learn > > > function, If someone sent me email that should be marked as spam I > > > would feed it with their username instead of the default username, > > > which is root. > > > > Yes, it will cause problems if you learn the forwarded message as-is. > > > > What you need to do is extract the attached email from the message > > (hopefully with the original headers intact), and feed it to SA as > > the user who originally received it. > > The problem is the original email is not attached, they just forward > it to me. Is there a way to just feed the body of the message instead > of the headers to bayes?
You could do something like that, but SA expects to get a full email. Feeding mail to bayes with no headers or with the wrong headers will cause unpredictable results. You need to train your users to forward suspected spams as attachments or create a shared folder they can copy them to. If the users' mailboxes are on the same machine that SA is on, you could create a this-is-spam type folder in each mailbox so they can just drop the messages there. Then you can have a cron job that regularly runs sa-learn on those directories and then deletes the messages (or copy the messages elsewhere if you want to maintain a spam corpus for rule testing). -- Bowie
