Hi, I am using SpamAssassin in a way I believe is different from most users. I have an IMail server (Windows 2003) which uses Declude JunkMail and a spamc client (spamc32) to connect to a spamd server on my Redhat FC3 box. It is working well in that I am getting spam reported by SA when invoking the spamc32 client from JunkMail. However, I am not certain that it is working altogether the way it should. I have trainied SA with both SPAM and HAM (several hundred SPAMs at least and a few hundred HAMs) but wonder how I can be CERTAIN that spamc32 is actually using the bayes database.
Maybe someone can tell me if this makes any sense at all: I create an email message and send it to a recipient and it is scanned by SA and found clean with a particular score. If I then feed that to sa-learn as SPAM and then resend the same message again, should it NECESSARILY have a different score, or is bayes looking at some many different things that it COULD score the same? In my test it scored the same. Here's how I am set up: I start spamd with a -u spamd user. When I run spamc32 I also run as this user. I don't believe the user piece is necessary in my case, I see no difference when I use it vs. not. I am trying to use a global setup, so all prefs are in the local.cf. There is also a user.prefs file in /root/.spamassassin, which is also where the bayes db is. Running spamassassin --lint -D does report it's using those locations. If it would help I can provide the whole output of that. I also don't see any additional info such as a header rewrite when a message scores 5 or more. Is this something that won't happen due to the way I'm using SA? I'm just not sure how I can verify what my spamc client is doing with spamd. Is it doing the basic default SA stuff and none of the stuff in my local.cf somehow? More info: SA version 3.0.4 local.cf: rewrite_header subject *****SPAM***** report_safe 0 use_bayes 1 bayes_path /root/.spamassassin bayes_auto_learn 0 skip_rbl_checks 0 use_razor2 0 use_dcc 0 use_pyzor 0 ok_languages all ok_locales all Thanks for any suggestions. Geoff
