On Sun, 2004-10-24 at 10:19, Chris wrote:
> 
> Peter, as shown below, network checks and the SURBL's have no problems 
> picking up the Rolex stuff:

Ah, but it is still useful for those of us waiting for a suitable moment
to upgrade from SA 2.64 to 3.0.1.  The time-line at my place of work is
the US Thanksgiving holiday.  While everyone else is away from work for
the holiday, I'll be shutting down the mail server and upgrading all of
the software, including SA.  Looking forward to using SURBL, but it will
be a few more weeks before I can make the switch.

The Rolex ads have been driving me crazy for the past week or so.  I
have written a number of ROLEX rules of my own, but they're pretty
crude.  I don't have much time available to devote to proper rule
design.  A failed hard drive mirror and overheated capacitors on the
motherboard have been more pressing problems in October. :-(    Bayes
has been really slow to recognize the Rolex spams, despite the fact that
I've been feeding them to sa-learn for nearly a week.

In the last month performance of my SA system has dropped from about 98%
of spam detected to only about 90% for some accounts.  I'm not sure why
the sudden increase in false negatives.

-- 
Fred W. Bacon <[EMAIL PROTECTED]>
Rigid Rotor

Reply via email to