I am currently running an OpenBSD mail filter in front of Lotus Notes with Postfix/Amavisd-new/SpamAssassin doing a decent job of filtering. I am preparing to make the move to Postfix/Dspam/ClamAV filtering and would like to train Dspam using my current setup.
The plan is based on this document: http://dspamwiki.expass.de/Installation/TrainingFromCurrent I've upgraded Amavis and SpamAssassin on the old box and will be adding the appropriate clean_quarantine and spam_quarantine lines to amavisd.conf. I've added the train-spam and train-ham aliases to /etc/postfix/aliases and run newaliases. I assume I'll need to add a row in dspam_virtual_uids for 'globaluser'? What about train-spam and train-ham? The article seems to indicate that dspam will consider them users since it recommends turning on Opt Out for them so they don't build a dictionary. We have multiple virtual domains which I map using the Postfix virtual table into one domain. Is there any need for me to setup [EMAIL PROTECTED], [EMAIL PROTECTED] and [EMAIL PROTECTED] all in the dspam_virtual_uids table with the same uid? Will dspam ever see anything besides [EMAIL PROTECTED] To rephrase, is virtual_alias_domain mapping done prior to content filtering in Postfix? One must hope so. Would it be easier and just as effective to load the SpamAssassin corpi (corpuses?) ? Thanks for your assistance. Michael D. Lemon Director of Information Technology Americal Corporation
