On Mon, Mar 08, 2004 at 07:11:00PM -0500 or thereabouts, Andrew Schulman wrote: > > I have Exim running, with spamc activated via Exim (not procmail). > > > > Now, when I'm training SA, I've been using 'sa-learn' as the user that > > reads the mail. However since Exim is run as mail user, should I be > > running 'sa-learn' as mail user? > > I'm not sure about this, since I call spamc from a filter in KMail, so it > runs under my UID. But it sounds like a FAQ. Exim can be configured every > way under the sun-- can you tell it to run a filter as the user who's > receiving the mail? e.g. call spamc -u $USER.
Hm, well I was thinking along those lines, but was looking for feedback from anyone using a similar setup. I guess there's no harm in trying it. Hell, it can't break anything. > > I'm wondering as I've been training SA for some time, and I'm not sure > > it's taking advantage of what's in my ~/Home .spamassassin directory, > > bayes_tok et al. > > A note about this: I was training SA for a long time, with several hundred > ham and spam messages via a daily cron job, and it didn't seem to be > learning. So I looked around, and found a note at the top > of /usr/share/doc/spamassassin/README.Debian to the effect that on > upgrading to perl 5.8, you have to delete all of your Bayes databases and > recreate them using sa-learn. I did this, and presto, SA started to work > like a charm. Apparently the default is 1k ham/spam b4 Bayes starts being accurate -- but I've got that amount. :) In terms of perl, I'm not sure that would be revelant in my case, as I haven't upgraded perl. It's a fresh install of the system. I kept my mail ham/spam intact though. Thanks Andrew. -- Steve +++++++++++++++++++++++++++++++++++++++++++++ Monday Mar 08 2004 11:06:02 PM EST +++++++++++++++++++++++++++++++++++++++++++++ If the meanings of "true" and "false" were switched, then this sentence would not be false.
pgp00000.pgp
Description: PGP signature