Most floating points aren't integers anyway, so if you think of it as casting 256 to the nearest u8 it's correct, same as rounding 0.25 to 0. NaN to 0 is a bit

Most floating points aren't integers anyway, so if you think of it as casting 256 to the nearest u8 it's correct, same as rounding 0.25 to 0. NaN to 0 is a bit more concerning, but inconvienence and compatibility need to be weighed against catching every error (no type system will catch every bug).