On Tue, 3 Feb 2004, Bob George wrote:

> You didn't really tell us HOW you use spamassasin.
> 
> Just as with a single site-wide bayes, there are possible problems
> adjusting site-wide rules, depending on what some users consider
> spam, or non-spam. I get lots of messages that fail one or more of
> the tests you mentioned, but that I still want for one reason or
> another. I'd be very annoyed if someone arbitrarily blocked them
> "for me".
> 
> So yes it might work fine, but you might also get some complaints
> if "semi-spammy" messages get tossed. What are you doing with
> flagged messages anyhow? If you're just tagging them, probably no
> big deal. I wouldn't set any one rule to make the final call in
> any case, unless this is a personal setup.
> 
> Are your users set up with personal configurations? Are you using
> site-wide bayes? Are you enforcing a corporate policy of some
> sort? Do you want "easy" or "accurate"?

Thanks for the reply.  I guess I'm looking for some sort of compromise
between easy and accurate.

I do NOT have a site-wide bayes setup.  My users can adjust their own
procmail setups allowing or denying specific addresses/subjects before
the messages get to Spamassassin.  They also have the ability to toss or
just mark spam.

My attitude toward some of the scores are that if a site is listed in a
spam database like spamcop, njabl or sorbs, that should be the end of it,
but I understand that a site can be used as a relay without the knowledge
of the webmaster, and can get listed before the webmaster can plug the
hole.  But knowing that I'm a neophyte at this, I'm sure it's not that
black and white.

--pat--
-- 
Pat Traynor
[EMAIL PROTECTED]

Reply via email to