PartialOrd documentation is misleading

The documentation for `cmp::PartialOrd` states that "if  `a < b` then `!(a > b)` and vice versa", but the "vice versa" is somewhat misleading here since `!(a > b)` doesn't imply `a < b` if e.g. either of the values is a NaN float. The author probably wanted to say that `a > b`  also implies `!(a < b)`.
