[H-GEN] Experiences with SpamAssassin

[Greg Black]

[ Humbug *General* list - semi-serious discussions about Humbug and     ]
[ Unix-related topics. Posts from non-subscribed addresses will vanish. ]

With the exploding volume of spam[1], I've finally decided to
give SpamAssassin a go.  Some of my experiences might be of
interest to others.

The installation sucks big time.  There's no "standard" way of
doing it and the default puts files in /usr/bin and /etc/mail
which should never happen.  I wrote a wrapper to use instead of
spamc so that, should I upgrade my OS (which would result in the
disappearance of SpamAssassin until I remembered to re-install
it), I won't just silently lose email.  But I should not have
needed to do that.  If there's a way around this suckage, it's
not documented in the obvious places and so can be said not to
exist.

To get an idea of performance (both speed and accuracy), I fed
spamc an archive of known email.  It took 14 hours to process
the test archive on a lightly-loaded Celeron-366 with 384 Mbytes
of RAM -- that's about 0.6 seconds/message.

Here are the details:

    Total archive size	84516
    Actual spam		 7918
    False positives	  218  ( 0.28% of good messages)
    False negatives	 1652  (20.86% of actual spam)
    Time to process	13:58

Some of the spam it failed to identify won't normally be fed to
it, but it's still going to miss about ten per cent of what I
get.  And, although the false positive rate of 0.28% is pretty
good, it's not good enough for me to just drop the identified
spam in /dev/null -- at least I can put it in its own folder and
review that folder once a day rather than getting constant
little announcements during the day about incoming email that
turns out to be spam.

Notes:

[1] My incoming spam figures for the past few years are:
        1998   358
	1999   506
	2000   448
	2001  1102
	2002  5586  (up to 22/11)

Greg

--
* This is list (humbug) general handled by majordomo at lists.humbug.org.au .
* Postings to this list are only accepted from subscribed addresses of
* lists 'general' or 'general-post'.  See http://www.humbug.org.au/