An excellent idea; I've been using precisely this mechanism for about a year now, except rather than an MD5 checksum I simply have a hard-coded constant password (same for everybody) that I ask the sender to respond with before I add them to my "whitelist".
(That's a good name for the "database" which Peter speaks of; it's the opposite of a blacklist.) Having a constant password is OK for the same reason that Peter mentions: spammers won't ever receive the mail. It works very well.
It's good to see that somebody has packaged this scheme up and made it available; my version is coded purely with procmail and a few shell scripts (no C++), but isn't really ready for distribution.
An open, cross-platform journaling program.
A scientific plotting package.