We use a single global Bayes DB. On that DB, we have months of learning
tokens. Even creating a new one does not help. This is an odd problem!
Justin Mason wrote:
aha -- that's being added by SpamAssassin alright, then, due to
the add_header line.
Are you using 1 global Bayes db, or per-user dbs? if the latter, maybe
there just isn't enough training for bayes to be active? Try getting debug
logs from SpamAssassin -- they'll fill you in on the reason (although I'm
not sure if MDaemon allows you to do that).
--j.
Dave Koontz writes:
Justin Mason wrote:
The only indication is SA's X-Spam-DataBase: Bayes not run. header.
SpamAssassin doesn't add a header like that... what are you using:
spamd, MailScanner, amavisd?
--j.
I use a product called MDaemon, which has a windows port of SA. Perhaps my
header mapping in local.cf is creating that header (see below). Either
way, there are no BAYES_XX hits on most of the messages processed. I
would say about a third are getting proper hits, all others have no BAYES
tag at all.
clear_headers
add_header all Report _REPORT_
add_header spam Flag _YESNOCAPS_
add_header all Status _YESNO_, score=_SCORE(0.00)_ required=_REQD_
tests=_TESTS_ shortcircuit=_SCTYPE_ autolearn=_AUTOLEARN_
add_header spam Level _STARS(*)_
add_header all DataBase _TOKENSUMMARY_
add_header spam RBL _RBL_
add_header all Hits _SPAMMYTOKENS_
