I'm setting up a new dspam installation (to replace an existing
amavisd-new + spamassassin one), and looking for some feedback /
recommendations on group setup.
My users are pretty much a 50/50 split between those that are technical
and I imagine will be good at reporting false positives / negatives etc,
and those that are not and will just expect it to work reasonably well
'out the box'.
Having reviewed the documentation and done some googling around etc, my
current plan is to have a global group, which I'm already training using
the output from my current filter (so it starts off around the same
level), and to also have an inoculation group based upon some honeypot
addresses which innoculates the global group user (unless I've missed a
way of having an innoculation group apply globally so it updates each
users dictionary without needing to define them all?).
My uncertainty is then around how, on an ongoing basis, I should train
the global group user with innocent mail (and spam, though the
innoculation group should cover most of this), as otherwise I imagine
its effectiveness will decrease over time, such that the users who want
things to work 'out the box', and new technical users who haven't built
up their own dictionaries will be disappointed.
This is assuming I'm understanding it correctly that a global group user
will be used to classify a mail, but not trained by it even if using
teft (I assume this must be the case as there would be on way to report
an error otherwise).
Any feedback on my approach would be much appreciated,
Director, Loho Ltd
Loho Ltd is registered in England and Wales number 7117766
Registered office: 141 Kings Hedges Road, Cambridge CB4 2PL