----- Original Message ----- 
> > I've had problems with learning spam that passed through
> > exchange. Due to
> > the fact that exchange changed too many headers, thus the
> > original spam
> > headers where lost. For example, exchange seemed to re-write
> > parts of the
> > email if it contained attachments and added a header like:
> > X-MimeOLE: Produced By Microsoft MimeOLE V6.00.3790.132
>
> this is one of the reasons I was asking...
>

That's not a problem:  Bayes will wind up ignoring any Exchange-specific
headers that occur in all mails.

> so, what you're saying is something like "unless you change your mail
> client, you can't feed spams to SA", is that it?
>
Yeah.  But create two public folders, SPAM and NOT-SPAM.  Hit it via IMAP
from say, Outlook Express, and use MailNavigator to convert the OE DBX files
to MBOX format on the spamassassin gateway, then sa-learn them.  Or even
more efficient, do it from Mutt or another IMAP client right on the
spamassassin gateway.  If the users drag & drop the mails into the public
folders, the headers should remain intact.  This works extremely well.

-tom

Reply via email to