> I believe "nan = math.huge - math.huge" is clearer. =) The point was not to depend on the math library.
