spamassassin & bogofilter

[David Relson]

Greetings,

Since almost the beginning, the bogofilter FAQ has said "How can I use 
SpamAssassin to train Bogofilter? (coming soon)".  Today I decided that 
it's time to fill in that hole.  So I loaded SpamAssassin-2.41 and did some 
procmail experimentation (along lines suggested by various posts on this 
list).  For testing purposes I used the sample-spam.txt and 
sample-nonspam.txt files that come with SpamAssassin.  At the end of my 
experiment, I had 3 test scripts, a procmail recipe, and some execution 
times.  Here's what I've got:

script #1 - tst.sa.sh - runs spamassassin twice - one time for each test 
file - to classify and rewrite it.

script #2 - tst.bf.sh - runs bogofilter four times - once for each test 
file to build the wordlists and once for each file to classify and rewrite.

script #2 - tst.pr.sh - runs procmail for each file.  The procmail recipe 
first runs spamassassin to classify the message, then runs bogofilter to 
register the message as spam/non-spam (depending on the results from 
spamassassin).  The script then runs bogofilter twice in order to classify 
and rewrite each of the messages.

In the name science, I used the "time" command to run the 3 scripts to see 
how long things took.  Here are the results:

tst.sa.sh	6.19user 0.28system 0:06.68elapsed 96%CPU
tst.bf.sh	0.16user 0.06system 0:00.49elapsed 44%CPU
tst.pr.sh	5.63user 0.19system 0:10.29elapsed 56%CPU

As can be seen bogofilter runs faster than does spamassassin.  No surprise 
there :-)

If you're curious, all files are available at 
http://www.osagesoftware.com/bf-sa/.

David