I have the following setup: 1) Firewall receives mail, pipes it through spamd and adds a header marking it as spam. 2) Firewall sends the mail through to internal machine. 3) Internal machine adds a 'Received:' line. 4) Internal machine detects that this is not spam but also not ham and wants the Bayesian filters on the firewall to FORGET this mail. 5) Internal machine bounces the mail to a special account on the firewall machine. 6) Firewall machine adds another 'Received:' header line. 7) Firewall stores the mail in an mbox called 'forgetmbox'. 8) cron processes the forgetmbox and executes sa-learn --forget on its content.
Question: Will the Basesian filters be correctly adjusted, despite the change of headers? What EXACTLY can I change in a mail so that it will still recognize the mail as 'already known'? What if a newline was added at the bottom of the body? Or when the mail would be (accidently) converted to another MIME encoding? I can imagine that it only looks at the Message-ID, unpacks the mail as far as possible to text and then reprocesses the words in it - in that case the header won't matter (except for the message-ID) and the (mime) encapsulation and extra new-lines etc won't matter either. But does it work like that? -- Carlo Wood <[EMAIL PROTECTED]> ------------------------------------------------------- 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