At the TiP BoF (98.5% awesome, btw), someone showed a tool for writing unit tests against a browser abstraction, and it could either use fake browsers (fast) or real browsers (accurate) using the same test code. What was that? Am I remembering it right? --Ned.