Hmm... I got this wrong, based on the BOL it's supposed to be FAST_FORWARD that is optimized for performance, but apparently, based on your paper, it's not. But I'm happy to lose the point as I learn something new again. I take my hat off to you for doing the testing.