The following is my take: It seems like the desired result is to make it determinable when a NaN value is produced in a language that doesn't have an is_nan fu

The following is my take: It seems like the desired result is to make it determinable when a NaN value is produced in a language that doesn't have an is_nan function (such as an assembly language lacking such a predicate) so that a simple boolean statement does the trick, for example: def is_nan(val): return not (val == val) This absolutely bizarre statement of non-identity gives us a surefire test. It is from the perspective of higher level languages that the non-identity of NaN with itself seems bizarre.