In message "[ruby-talk:03133] Re: Is divmod dangerous?"
    on 00/06/06, Dave Thomas <Dave / thomases.com> writes:
> >It seems as if changing division so that 13/-4 -> -4 is a popular
> >option, perhaps keeping the current behavior as a new method. a%b
> >would then become a synonym for a.remainder(b).

>I guess it might just be me that was surprised to find that
>
>    q, r = divmod(x, y)  =/=>  x = q*y + r
>
>(where =/=> is a lame attempt at "does not imply"). In an ongoing
>attempt to minimize the surprise function, I thought I'd raise the
>issue.

Oh, I misunderstood :-(  Even in C, div(x,y) satisfies 

  x == y*div(x,y).quot + div(x,y).rem

Okay, divmod should be change.  I vote 13/-4 -> -4 too. 
Generally, 

  num / den := floor(num "div" den)  # "div" in done in the real
  num % den := num - den*(num/den)

-- gotoken