* Marc Perkel <marc@???> [2004-01-02 17:02]:
> That's a good idea for most people. In my case I want to actually
> accepts the spam - bayes learn it - report it to Razor - and then bounce
> it as an non existent address.
Please dont't do that!
If you _know_ it's spam then don't ever send a separate bounce email
after accepting the message because this will annoy a lot of innocent
third parties whose sender addresses were forged in the spam. Today most
spam uses forged envelope sender addresses.
Instead reject it at smtp-data time (using exiscan-acl or sa-exim - this
way could also include bayes_learn or razor-report) or if already accepted
just tag it, sort it in separate spam-folders or (if the users are informed
and agreed to) just drop it.
