From: Jorey Bump (no email)
Date: Thu Dec 01 2005 - 16:07:04 EST
Covington, Chris wrote:
> this constantly gets better): The database is specially-tailored to
> our users' email patterns, and our users continually update it themselves
> by forwarding false positives and negatives to training addresses.
> So if you have the patience, skills and hardware required to use
> DSPAM, go for it!
Maybe I'm missing something, but a system that requires users to
continue to handle spam (if not actually read it) *and* to learn another
interface to train the application seems like little more than a mail
sorting program to me.
I want to reject spam immediately, during the SMTP conversation. I'm
doing that now on a low-volume site, running SA in a before-queue
content filter using spampd. While it may be possible to do the same
with dspam, SA bootstraps its bayesian filter with a multitude of rules,
so no user intervention is required. For the little bit of spam that
gets through, I can run sa-learn, or, even better, add a local rule that
will improve the bayesian filter after a few rejections.
|
|
|