> Message: 7 > Date: Fri, 25 Oct 2002 11:08:39 -0400 (EDT) > From: Ed Hennis <[EMAIL PROTECTED]> > To: [EMAIL PROTECTED] > Subject: Re: [Razor-users] Re: Razor-users digest, Vol 1 #404 - 13 msgs > > But what if, hypothetically, my filters are never wrong? Hypothetically, > I've saved you the need to use your CPU cycles to run the filter yourself. > So it is to help; not only you, but every user of razor, including those > who are not experienced, smart, or savvy enough to know or install those > other filter programs.
If it's never wrong, then I'd be happy to waste the CPU cycles on my own machine. If the CPU cycles were worth saving, I'm sure that a razor-like database could be implemented, parallel to Razor proper, for the dissemination of the decisions of this "perfect" filter. Perhaps the Razor developers could even be convinced to bless the filter with a special permission to autoreport, maybe even with a fixed 100% trust score, thus enhancing Razor's effectiveness. However, I don't trust any old programmer/sysadmin's judgement that their automated system generates "sufficiently few" false positives that it would be beneficial to pipe its raw output right into Razor. > "The avalanche has already started. It is too late for the pebbles to > vote." > -- JMS, via Kosh :-) acknowledged. I can see the evidence in /tmp/spam-razor... > The plain and simple fact is that if the _software_ doesn't stop the user > from doing something, they're probably going to do it. And the funniest > part about this whole argument is that the trust system is essentially > _already_ stopping the user from doing the "wrong thing". I understand that, and the developers do seem to share my view that a false positive should hurt your rating more than a true one helps it. However, for the past week I have gotten at least one false positive a day out of Razor2, which is probably about 20-30% of the actual spam it catches. I am using multiple DNS based blacklists to reject spam before it goes to Razor, so that increases the proportion of false positives, but if the ratio gets much worse I'll probably take razor out of the sequence. Solutions? I don't really know. Maybe increase the asymmetry between the penalty for bogus reports and the reward for good reports (an autoreporter could potentially make hundreds of good reports for every false positive, especially if it's upstream from razor-check). Or razor-report could be written so that it takes no action if it is called by procmail or other common automatic mail processing programs, forcing the errant sysadmin to be less "lazy" if they really want to buck the system. -- [EMAIL PROTECTED] John Stimson http://www.idsfa.net/~john/ HMC Physics '94 ------------------------------------------------------- This sf.net email is sponsored by: Influence the future of Java(TM) technology. Join the Java Community Process(SM) (JCP(SM)) program now. http://ads.sourceforge.net/cgi-bin/redirect.pl?sunm0004en _______________________________________________ Razor-users mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/razor-users
