Having 1/0 = 0 and 1.0f / 0.0f = +inf makes sense in two applications I can think of. The first would be currency systems in which prices down to the pre-determined fractional component are represented as unsigned integers. In calculating margin requirements division by zero equating to zero would be the correct answer. And two, for double precision continuous variables, any application involving 3D geometrical transforms. An aspect ratio approaching zero in one axis, would approach positive infinity in the other.

So you may be saving some developers time in operator overloading, etc. More important than the actual binary decision itself is having well-documented behavior. Have not tried Pony yet, but will take a look. Good discussion, keep it up!