On Wed, Sep 10, 2003 at 04:03:02PM -0700, John Schneider wrote: > This seems to work OK, but I'm not sure it is an efficient setup.
It is not. Provided that you have a scenario with non-linear delivery and/or multiple users, there is practically no use at all in calling spamassassin when you could call spamc to connect to a running spamd. spamd is SpamAssassin's core classes preloaded and precompiled held in memory. It is also a kind of load management, if you are using the --max-clients (-m) parameter properly. The spamc is a very lightweight client that passes the mail to your spamd and has is analyzed in that preloaded and precompiled perl. That setup is much more efficient than a single spamassassin (except in very few pathological cases that you do not need to care about). Particularly, using spamassassin from procmail, in multiple users procmail configurations, has no load control at all. There is no limit on the number of parallel, concurrent spamassassin startups on your machine at all, and a mail storm (for example after some user has fetchmail'ed a hundred mails to his local account) will bring down such a setup completely. On the other hand, this is not a problem at all for spamd with a proper -m switch. spamc is more or less a drop in replacement for spamassassin, and deploying it should be a simple s///g for you. > Also, I have SpamAssassin 2.55. I see a lot more spam has > found it's way through the last week or so. Is there an > evolving set of rules that deal with this that I should be > updating? Use Bayes and AWL. This is a pretty good method to keep SpamAssassin sharp and up to date with many new spams. My own setup, which is perhaps overly restrictive for most people, looks like this: ---- .procmailrc ---- PATH=/bin:/usr/bin:/usr/local/bin MAILDIR=$HOME/Mail LOGFILE=$HOME/Procmail/procmail.log DEFAULT=/var/spool/mail/kris BOUNCE=/dev/null SPAM=$MAILDIR/spammed :0fw | /usr/bin/spamc # Tons of mailing list rules cut out :0 * ^X-Spam-Level:.*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\*\* $SPAM-record :0 * ^X-Spam-Level:.*\*\*\*\*\*\* $SPAM-sure :0 * ^X-Spam-Level:.*\*\*\*\* $SPAM-probable ---- .spamassassin/user_prefs required_hits 4 whitelist_from <some friends> # So you can sort by subject and get mail sorted by Spam level subject_tag SPAM _HITS_ - rewrite_subject 1 # I do not take either score MIME_HTML_ONLY 4.0 score MICROSOFT_EXECUTABLE 4.0 # This is what I read and write ok_languages de en ok_locales en # bayes use_bayes 1 auto_learn 1 # very restrictive, and not generally recommended auto_learn_threshold_nonspam 0 auto_learn_threshold_spam 6 # Razors use_razor1 1 use_razor2 1 ----- Kristian ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
