ComputerWorld has an opinion piece about testing:
Here's the detonator: If a new release adds just 10% to the existing code base, the amount of testing that's required to be sure the new stuff works, and everything that used to work still does, is 110% of the previous release -- yet the time and resources applied to testing all of this functionality are at best flat release over release, and more likely declining. And, it was never more than a fraction of the development effort to begin with.
I liked her position that semantics does matter, and that we should possible call regression testing something else. Her idea of "operational assurance" is just as good as anything I can come up with.
Of course, with testing tools like Team System, we can always make sure we cover our bases! (Had to put the plug in there!)