> That I can attach more information to the result---specifically, the > source-location information can be a lot more precise that way. Fair enough. Matthew and I discussed this very point at Snowbird, and I agreed to go with the entirely opaque model, albeit with some trepidation about the loss of convenience. Kent