Installation instructions for SpamAssassin 3.1.3
Laird Bedore
7/19/06

   ## Required Reading:
   ##  http://svn.apache.org/repos/asf/spamassassin/branches/3.1/README
   ##  http://svn.apache.org/repos/asf/spamassassin/branches/3.1/INSTALL

   ## This document depends on the qmail+spamcontrol 2.4 installation
   ##  instructions written by Laird Bedore being completed beforehand.
   ##  It assumes that your mail server is already running and tested.

   ## Get the prerequisite perl modules from CPAN.

   ## You'll need lynx at the minimum in order to use the "cpan" command 
   ##  to grab the modules. If you don't use "cpan", you're crazy.
    % swaret --install lynx
    % cpan
   ## Answer the myriad questions. Defaults are fine for just about everything.
   ##  Now, in cpan:
    cpan> install Digest::SHA1
    cpan> install Digest::HMAC
    cpan> install MIME::Base64
    cpan> install HTML::Tagset
    cpan> install HTML::Parser
    cpan> install Net::IP
    cpan> install Net::DNS
    cpan> install Time::HiRes
    cpan> install URI::Escape
    cpan> install Net::CIDR::Lite
    cpan> install Sys::Hostname::Long
    cpan> install Mail::SPF::Query
    cpan> quit

   ## Unpack the tarball
    % cd /usr/src
    % tar zxf /vsn/pub/qmail/spamassassin/Mail-SpamAssassin-3.1.3.tar.gz
    % cd Mail-SpamAssassin-3.1.3
    % perl Makefile.PL
   ## You will be asked some questions. Answer them!
   ##  URL for more info: http://www.vectorstar.net/spamassassin.php
    % make
    % make install
    % cd spamc
   ## We don't actually need this, but lets build it in case we use it in the 
   ##  future.
    % make qmail-spamc
    % cp qmail-spamc /var/qmail/bin

   ## in /etc/mail/spamassassin, edit the config files and tune to your liking.
   ##  Be careful to get these things right, after all.
   ## In v310.pre, make sure the AutoLearnThreshold is enabled and AWL (auto
   #   whitelisting) is disabled.
   ## Settings in local.cf:
	rewrite_header Subject (SPAM Score: _SCORE(0)_)
	report_safe 0
	lock_method nfssafe
	required_score 8.0
	#   Bayesian Filter settings
	use_bayes 1
	bayes_auto_learn 1
	bayes_auto_learn_threshold_nonspam 1.0
	bayes_auto_learn_threshold_spam 12.0
	bayes_min_ham_num 25
	bayes_min_spam_num 50
	# max tokens in db: default 150,000 - approx 8mb
	bayes_expiry_max_db_size 75000
	bayes_auto_expire 1
	bayes_ignore_header X-Bogosity
	bayes_ignore_header X-Spam-Flag
	bayes_ignore_header X-Spam-Status
	#   Other settings
	report_hostname mail.vectorstar.net

   ## Configure spamd for daemonized operation. universal unpriv mode or
   ##  as-each-user mode? Need to test more!
   ## Qmailrocks.org has an example script for launching spamd:
   ##  http://www.qmailrocks.org/clamspam_slackware.htm
    % cp /vsn/pub/qmail/spamassassin/rc.spamd /etc/rc.d/
    % chmod 755 /etc/rc.d/rc.spamd
    % echo "/etc/rc.d/rc.spamd start" >> /etc/rc.d/rc.local
    % /etc/rc.d/rc.spamd start

   ## Create special configuration for system-wide bayesian filter:
    % cd ~qmaild
    % mkdir .spamassassin
    % pico .spamassassin/user_prefs
	# Local spamassassin configuration file for qmaild (used on systemwide scans)
	use_bayes 1
	bayes_min_ham_num 100
	bayes_min_spam_num 200
	# 1 million tokens - about a 50mb db
	bayes_expiry_max_db_size 1000000
	bayes_auto_expire 1
    % chown -R qmaild .spamassassin
    % chmod 700 .spamassassin
    % chmod 600 .spamassassin/user_prefs
    % chgrp nofiles .spamassassin/user_prefs
   ## SpamAssassin will automatically create and maintain the files 
   ##  "bayes_seen" and "bayes_toks" within ~qmaild/.spamassassin/

   ##  Edit your qmail-qfilter script and run "/usr/bin/spamc"..
    % pico /var/qmail/bin/prequeue-filter
   ## It should have an execution line like this:
	exec /usr/local/bin/qmail-qfilter /usr/bin/spamc

   ## That's it! Everything is set up and running. This provides a system-wide
   ##  spamassassin filter with a reasonable 8.0 score requirement for spam.
   ##  All email, regardless of score, is delivered. System-wide Bayesian 
   ##  filtering will be done  once enough emails (100 good, 200 bad) have
   ##  been received and learned by the bayesian filter.
   ##
   ## How to configure per-user bayesian filtering:
   ##  To turn on per-user filtering, create the following .qmail file
   ##  in the user's home directory with the correct ownership/permissions:
	|/usr/bin/spamc -e /vsn/bin/mailfilter
   ## SpamAssassin will automatically create and maintain the bayesian DB.
   ##  To populate the DB based on the user's emails, run the following
   ##  commands as the user:
    % sa-learn --ham --mbox ~/.mail/INBOX
    % sa-learn --spam --mbox ~/.mail/spam

   ##Check the following:
   ##  send some test emails.
   ##  Make sure the emails have arrived. If not, you probably need to
   ##   revert the qfilter changes and read the qmail logs in /var/log/qmail
   ##   and /var/log/qmail/smtpd .
   ##  check that ~qmaild/.spamassassin/bayes_seen and bayes_toks have been
   ##   created.
   ##  Read the messages. Look at their headers and make sure that the 
   ##   X-Spam-whatever headers have been added by SpamAssassin.