On Monday 06 October 2003 10:41 am, Praedor Atrebates wrote: > I have fiddled some with my spamassassin but it is still just too slow. I > am running it in daemon mode (spamd and spamc) with bayes filtering on. > Spamassassin email analysis on my laptop is too slow to live with (about 20 > seconds per message). Several questions with the first: How do I add a > mailing list like this one to my whitelist? The whitelist appears setup to > deal with "from" addresses, not "to" addresses. Since all list mail is > "To: [EMAIL PROTECTED]" or some other variation of this, I can't > see how to setup my whitelist for this.
A lot of this depends on what you consider "slow". My own setup takes about 40-60 seconds per message for filtering, I figure a lot of that is from the RBL checking, I use RBL lists extensively and some of them are a little slow to respond to queries. One suggestion, if you know that you want to whitelist mailing list messages, why send them through spamassassin at all? Simply setup a procmail recipe to put those messages directly in your spool, that way, you don't waste time doing header and body checks on messages that you would have whitelisted to reduce the score of those checks on anyway. In short, why do the checks if your whitelisting score will reduce the score to the point that the checks are superfluous? Something like: :0H * ^List-Owner: <mailto:[EMAIL PROTECTED]> /var/spool/mail/user will put all Mandrake Expert messages into user's mail spool and will essentially bypass spamassassin if that recipe is used first, before the one that sends all remaining mail to spamassassin. > My next question is about how to > reduce spam analysis time down from 20+ seconds to something more > realistic. Here are the entries to my user_prefs file, exluding whitelist: > > score RCVD_IN_RBL 10 > score RCVD_IN_RSS 1 > score RCVD_IN_DUL 1 > score RCVD_IN_BL_SPAMCOP_NET 4 > > use_bayes 1 > score BAYES_00 0 0 -6.400 -6.400 > score BAYES_01 0 0 -6.600 -6.600 > score BAYES_10 0 0 -6.400 -5.801 > score BAYES_20 0 0 -5.801 -3.101 > score BAYES_30 0 0 -1.246 -1.604 > score BAYES_60 0 0 2.002 2.002 > score BAYES_70 0 0 2.637 2.637 > score BAYES_80 0 0 4.1 4.1 > score BAYES_90 0 0 4.2 4.2 > score BAYES_99 0 0 4.300 4.3 > > What should I add or remove from the above to make this work faster? It might help if you told us what you consider "realistic" to be. I would suggest removing the RBL checks altogether if you really want to speed things up. With those, you are always going to be limited by the speed of the Internet and the DNS servers you are querying. Even with RBL timeout set relatively low, I think that it is cumulative, so a 10 second time out with 4 RBL checks adds up to a potential 40 second delay. -- Bryan Phinney Software Test Engineer
Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com
