Andrew Sutton

Well, no... Division by 0.0 _is_ defined, and the result depends on the dividend. I think Sean's point (it's definitely mine) was that there are cases of difficult-to-specify sets of...

> IEEE-754 floating point numbers are not the real numbers. Your > misinterpretation of my statement does, however, point out that division by > zero could be a poor choice...

> Where does this discussion leave us? Does double satisfy > EqualityComparable? If yes, what do we say about outlier values like NaN? > If no ... I'm not sure...

We need a way of systematically documenting which operands would result in undefined behavior. We can do this on a concept-by-concept basis for built-in types and other general sets of...