Domino/Exchange MailScanner users

[Pete]

I know this has been covered before, but we are a Domino shop with
MailScanners running in front of our email borders. All non spam email
is relayed to Domino servers.

It is not possible in this environment to ask users to forward mail to
spam/non spam addresses, or copy them to spam/non spam folders etc they
just wont do it.

Is it practical for me to archive some non spam email each day and some
spam email then run sa-learn over it with Julian's bash scripts?

Is this how people who dont have the spam/non spam boxes gather non spam
for bayes? I keep trying to use autio learn but i dont think it work
very well and is very quickly poisened and giving negative scores to spam.

Not using bayes means too many newsgroup/newsletter emails are trapped
as spam, being an academic facility i, like mailscanner, consider most
of the emails these people get is spam, but they dont. So i need to put
some effort into getting bayes working, but without any user input.

1. Is the abiove worthwhile persuing? or does bayes really need user to
input spam/non spam?
2. If i archive non spam and feed it into bayes, i would need to have a
good look at it first - is there a way to apply filters to mailwatch so
that mailwatch will display say only Spam, or only High Spam, or only
emails that were archived, or only Non Spam?

With this filtering i could then check the mail each day easily and
release the legit stuff and then run the scripts on the mail remaining.

Thanks in advance for ANY suggestions.
Pete