On Wed, 13 Jul 2005, Henry Cejtin wrote: > I would say that the change to stop Real.real from being an equality type was > not because testing against 0.0 is probably foolish, but because of the fact > that IEEE equality is NOT reflexive. If z is NaN then z = z is false. That may have played a role too, although other languages are quite content to use z==z as the test for non-NaN-ness. Peter