Am 23.06.2015 um 18:48 schrieb Bill Cole:
***** sa-learn IS NOT THE RIGHT TOOL FOR LEARNING MESSAGES INTO A SYSTEM-WIDE DB ****
says who? that below is the rsult of a customized sa-learn script for a ton of users working like a charm on a spamass-milter setup for 10 months now
[root@mail-gw:~]$ bayes-stats.sh 0 30009 SPAM 0 17486 HAM 0 2125684 TOKEN insgesamt 64M -rw------- 1 sa-milt sa-milt 5,0M 2015-06-23 18:16 bayes_seen -rw------- 1 sa-milt sa-milt 80M 2015-06-23 18:16 bayes_toks -rw------- 1 sa-milt sa-milt 98 2015-02-17 11:37 user_prefs BAYES_00 27854 71.55 % BAYES_05 856 2.19 % BAYES_20 955 2.45 % BAYES_40 857 2.20 % BAYES_50 3398 8.72 % BAYES_60 359 0.92 % BAYES_80 390 1.00 % BAYES_95 298 0.76 % BAYES_99 3962 10.17 % BAYES_999 3654 9.38 % DELIVERED 44487 89.75 % DNSWL 41562 83.85 % SPF 19953 40.25 % SPF/DKIM WL 10265 20.70 % SHORTCIRCUIT 10617 21.41 % BLOCKED 6154 12.41 %
signature.asc
Description: OpenPGP digital signature
