> -----Original Message----- > From: Robert Menschel [mailto:[EMAIL PROTECTED] > Sent: Thursday, August 07, 2003 11:12 PM > To: Chris Santerre > Cc: [EMAIL PROTECTED] > Subject: Re[4]: [SAtalk] SA tweaking (was: some spam subject header) > > > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Hello Chris, > > Thursday, August 7, 2003, 7:04:58 AM, you wrote: > > RM>> - From my point of view, spam detection should be handled by > RM>> 1) distributed rules (with a few score modifications if > appropriate) > > CS> Yup. Adjust to taste. > > RM>> 2) DSNBLs > > CS> But some may not be able to use. I unfortunetly have to > do business > CS> with countries that are spam havens. I can't afford the FPs on a > CS> business level. > > What problems do you run into? With which countries? I'm > handling a fair > amount of business email from Asia and Mexico and haven't noticed any > problem. >
I deal with Korea (kornet) which is just spam crazy. Chinanet I wish would dry up and die :) I've heard many a RBL block these. I can't take a chance, as the emails we get are time sensative. Missing 1 good email could really suck. > RM>> 3) Bayes > > CS> Some may not be able to use, or use easily. This is due to server > CS> setups. I'm one of these at the moment. I will attempt w/ 2.60 SA. > > My sympathies. Don't worry, I'll get to it...soon. :) > > RM>> 4) blacklists (I haven't praised William Stearns' collection > RM>> in the last > RM>> five minutes, have I?) > > CS> I believe these should be at the lower level before SA. > Why waste the > CS> cycles :) > > Because some of us may not be able to use, or use easily. :-) As an > end-user on a virtual domain server, I don't seem to have > direct access > to procmail or other methods of implementing blacklists before SA. > Blacklists within SA are functional and well worth while for me. Ah! My own medicine! I see your point. > > RM>> 5) personal rules > > CS> Which leaves me with basically 1 and 5. :) > > CS> Don't get me wrong, I was excited about Bayes! Still am. > But because > CS> of my setup, I had to put it aside. ... > > CS> I figured others might be in same boat. Admins not > wanting to explain > CS> to users how to feed bayes. ISPs not wanting to use > DNSBLs. ect.... > > I really don't see a need to explain to users how to feed bayes. My > server provided an SA on/off switch back in May, and I turned > it on for > one of my domains. I liked the results. Before I knew it, Bayes had > trained and was trapping spam for me. auto-learn is quite > sufficient on > its own as far as I'm concerned -- feeding bayes specific emails is a > nice-to-have feature which can give an incremental benefit, > but it's not > necessary. > So how would an end user fix an FP? What if autolearn hit an FP, which increased into more FPs. Could snowball without fixing. I'm talking without a clue, as I still haven't got to bayes yet, so I might just be talking silly :P Chris ------------------------------------------------------- This SF.Net email sponsored by: Free pre-built ASP.NET sites including Data Reports, E-commerce, Portals, and Forums are available now. Download today and enter to win an XBOX or Visual Studio .NET. http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01 _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
