Simon Byrnand writes: >Just browsing through my bayes database and I see > >debug: bayes token '8.11.6' => 0.0186046511627907 > >Presumably it has grabbed the 8.11.6 from our sendmail version which is in >all message headers processed by our server ? > >Which begs the question, why has bayes learnt that 8.11.6 is a ham >indicator when it is present on EVERY message it ever sees on our system, >ham or spam ? :) Shouldn't it gain an indeterminate score somewhere in the >middle and thus not contribute to a bayes score ?
There's a possibility that most of your spam is coming via a secondary server running some other version.
Nope, only one secondary, and it runs the same version of Sendmail. (Changed to 8.11.6p2 a while back, but slightly after the bayes database was already learning)
Alternatively, what's the token counts for that token? And what's the nspam and nham counts at the top of the dump?
Ok, well that first line as you probably guessed was taken from scanning an (older) spam in debug mode, so here is the relevant output from check_bayes_db, including all matches for things similiar to 8.11.6
0.000 0 0 0 non-token data: db format = on-the-fly probs, expiry, scan-counting
0.000 0 28070 0 non-token data: nspam
0.000 0 44894 0 non-token data: nham
0.000 0 102780 0 non-token data: ntokens
0.000 0 14341 0 non-token data: oldest age
0.000 0 53621 0 non-token data: current scan-count
0.000 0 52516 0 non-token data: last expiry scan-count
0.035 0 6 53544 8.11.6 0.666 23617 18961 53632 H*r:8.11.6p2 0.011 2 302 53544 8.11.6p2
Regards, Simon
------------------------------------------------------- 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/psa00100006ave/direct;at.asp_061203_01/01 _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
