Setting up dspam is pain in the ass. It is too complex and there is way too much space for bugs, many of them may cause silent mail loss.
There is one tool developed exactly for what dspam does -- milters. Setup is really simple and there is almost nothing to screw up. Just name the socket, make sure it has correct permissions and dspam is ready to go.
For example opendkim uses this to validate signatures.
Setup in postfix looks like this (main.cf):
smtpd_milters = unix:/milters/opendkim milter_default_action = accept
The first line specifies socket for opendkim and second says that if opendkim is dead, mail should be delivered anyway (default is to hold it in queue, which is fine too).
Why dspam does not use this milter api too?
I found some dspam-milter projects in python, but it looks like one big hack and far from bulletproof.