Is this the correct command sequence for starting and stopping filters? I have one filter, a perlfilter.

# filterctl start perlfilter # courierfilter start # filterctl stop perlfilter # courierfilter stop

On a couple occations, I have ended up with processes that I could not get rid of by using filterctl stop or courierfilter stop.

root@spaminator# courierfilter stop root@spaminator# ps ax | grep filter 24039 p2 IJ 0:00.00 /usr/local/sbin/courierfilter start 24041 p2 IJ 0:00.00 /usr/local/sbin/courierlogger courierfilter 29166 p2 S+J 0:00.00 grep filter root@spaminator# kill 24039

I assume I am just running commands in the wrong order.

Second, I had an instance where one of the razor perlfilters got borked somehow, and was not able to find the server to lookup the message hash. The other perlfilter instances where chugging along just fine. I'd like to protect against this somehow with cron--is it safe to kill a perlfilter process? Or should I be conservative and just do the following when cron sees a borked perlfilter:

# filterctl stop perlfilter # courierfilter stop # ps ax | grep filter | awk '{print $1;}' | grep -v PID | xargs kill -9 # filterctl start perlfilter # courierfilter start

Also, the dupfilter example that comes with Courier is a baby Vipul's Razor! Is there any interest and/or expertise here in helping dupfilter grow up and learn how to collaborate with others? Looks like the hard bits are availability of central hash database (p2p maybe?), computing message hashs in a smart way (really hard I think), and implementing trust metrics of reporters. Although Mr. Prakash is a staunch Free Software advocate, I would prefer an alternative where the hash database and trust algorithm was not proprietary. A periodic rsync of the central db to a local db and a courier filter in C would be nice and fast. :)

m