From: "Marc Perkel" <m...@perkel.com>
Sent: Wednesday, 2010/February/03 09:20

jdow wrote:
From: "Alex" <mysqlstud...@gmail.com>
Sent: Monday, 2010/February/01 11:24

That's a bad thing for anyone, not just hospitals, but I doubt if the
system that sends regular email is in any way connected to the
internal patient system.

Not knowing what their system is I have to make sure that email sent from hospitals gets delivered. Passing ham takes precedence over blocking spam.

Yes, agreed; I just wanted to point out to jdow that the internal
systems are much different than their public systems, so a compromise
of their public system doesn't necessarily mean patient records are at
risk.

It creates at least a perceptual problem, I believe.

{^_^}

I don't see it as a perceptual problem. What rules are to lower the score of ham. SA really needs more white rules. White rules can compensate for the sins of black rules and enhances overall accuracy especially when protecting ham take priority over blocking spam.

In that regard I quite agree with you, especially since I mark up the
Bayes99 score to 5.001.

Meta rules come to my assistance there. I've isolated 99% of the mis-
scored email problem to mailing lists. Almost all their email is clean.
But some lists are cleaner than others. And in many lists I am on the
Bayes scores tend to be random but below Bayes80. So I use meta rules
to refactor Bayes scores around 80, increasing the spam score above
Bayes80 and reducing it below Bayes80. That way my other rules which
are also triggered by say the "random text gibberish" in the LKML are
neutralized while the spam scores are actually enhanced.

It requires care and feeding. (And, grin, Ubuntu has different rules
from Fedora and LKML.)

Regardless, for my needs I'd never run a HOSTKARMA_W score as far
negative as -5.0. Lately I ran across one escaped spam that was
particularly obnoxious to find in my clean mailbox.

{^_^}

Reply via email to