[thaumaturgy]

Arin will allow you to download portions of their database for research purposes, but it requires an account and it looks like there's an approval process.

You wouldn't have to query every IP, you'd just have to query IPs not covered by a result range from a previous lookup.

I'd be willing to help on this; I have a whois module I wrote for another spam tool, I should be able to pretty easily adapt it for this.