> And computers still don't really understand either what issue you're looking for or what issue a case is about. So ancient technology (search for this word near that word) still rules the day.

I think this is the key. The big revolution in legal research that I'm waiting for is the ability for the computer to understand something like: all cases where a) Claim X was brought as a counterclaim and not as the original claim; b) counter-claimant made Y argument citing to case Z but not case W; d) requested relief was R; and e) court made its decision based on Factor F. Until then, "search for this word near that word (and boolean operators)" is extremely powerful and can get me most of the way pretty quickly.