[JoshTriplett]

The compiler doesn't necessarily have to have that level of inference; it can simply require the code to prove the constraints. For instance, parseInt will return an unconstrained int, so you'll get a type error trying to assign that to a range-constrained int. For your first and last examples, the multiplication operator on range-constrained ints can easily propagate the range constraints. In your last example, `z = w * x` would work, but assigning through the unconstrained int y may simply cause the last assignment z=y to produce a type error, forcing the code to re-establish the constraint explicitly.