I think I would second everone else's suggestion that you need to test-run
this and see. There is Matt's stock antidrug stuff, and as you note we have
a few more SARE rules of our own.
I know that none of us SARE types have access to a corpus with any
significant legit medical traffic, so I'm sure that we wouldn't certify our
current scores (or even rules and combinations) as safe in this environment.
We would say that if you use descretion and are willing to tweak the scores
you could probably use many of the drug rules though.
What would perhaps be most interesting, if you have a way you could
configure it, would be to set up a test SA install that would test-classify
the incoming mail stream using the currnet rules and scores. That way you
could start seeing immediately what isn't working for you and needs
tweaking. You could then block out the bad rules for your real SA install
and keep the others.
Of course, we'd really like to know what doesn't work for you so that if
needed we could modify the rules. Possibly adjusting things to use some
metas might reduce the FPs you see (assuming you see any at all) and still
keep the good properties.
If nothing else, I'm sure we would be willing to publish a scoreset or
version of the rules that you have found to be workable in a medical
environment.
Loren
