David Crossley wrote:
Stefano Mazzocchi wrote: <snip/>
But my biggest concern is about false positives.
One solution would be to use spamassassin for tagging purposes only, but at that point it's much better to let people do the filtering themselves. There is no reason in wasting precious CPU power for that.
A more socially-acceptable compromise would be to leave the "threshold level" of the cutoff point to the various people to set (you could set a .spam_threshold file in your home directory with the cut-off point). So, if you keep the threshold low and complain, well, that was your own fault.
I would prefer to see server-side filtering because i don't want to even receive the volume of junk. It is a big waste of precious bandwidth.
True, but keep in mind that you can still do server-side filtering, just you have to do it on your own (as I do already today, for example).
Surely we can tune SA to minimise false-positives, especially since we have the experts themselves at Apache.
Being an expert in pattern matching doesn't make you an expert in understanding what is spam or not for somebody.
My point is that no matter what, there is only so much you can do on a centralized spam filter and you should not, IMO, attempt to do to much.
That's why I think leaving the threshold configurable by person is the best approach since it leaves you the ability to "turn the nob" in how selective your filter is.
-- Stefano.
smime.p7s
Description: S/MIME Cryptographic Signature
