[Mimedefang] MIMEdefang and Spamd/Spamc
skmimedefang at smail.inf.fh-bonn-rhein-sieg.de
Mon Nov 12 03:25:11 EST 2007
-----BEGIN PGP SIGNED MESSAGE-----
On Sat, 10 Nov 2007, Rick Knight wrote:
> getting error messages about "malformed_spamc_report". It seems the spamc
> report is not being read by mimedefang-filter. Can any one offer a suggestion
> on how to correct this? Or maybe another solution?
> I'm running MIMEDefang 2.63, SpamAssassin 3.23 and CLamav 0.92 on Slackware
> with Kernel 2.6.14.
I run plain Debian Etch with spamassassin 3.1.7-2 and run such setup
since Sept 2007.
I have adopted a Client.pm that is shipped with SpamAssassin:
I added some checks, the possibility to push contents from a file (rather
than from memory) and tweaked the API so it returns what I think is
sensible. Also included is the part in my filter. (Maybe I cut off too
much, but the idea is readable.)
Overall, I cannot say that by using spamc the system load or performance
got better, also: spamc forks each time a request is passed to it, but the
shared memory is very small: For testing purpose, I configured spamc to
use William Stern's Blacklist and some other ones, each process was approx
400MB in size, but the individual clients scanning a request have only
636KB shared memory. Maybe I interprete the memory statistic wrong, but it
doesnot look nice.
David already mentioned that this behaviour is (most likely) related to
the reference counting of Perl's data management.
I did went the route to spamc.pm mostly, because I use SPAM checking also
from outside MIMEDefang, but I start to wonder if I revert the step.
But there is one advantage: If you can trust the local socket of spamd,
you can run spamd as root and can have spamc pass over any user you like
to have spamd read its configuration from.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
-----END PGP SIGNATURE-----
More information about the MIMEDefang