Re: operation of 'round' function

[Ronan Waide]

pjacklam@online.no (Peter J. Acklam) writes:
> I guess this proves the often-cited text in the Perl
> documentation.  It is equally true for Emacs.

Having received a few emails and apparently opened something of a can
of worms on the topic:

* I'm aware that my initial assumption was incorrect, based on a mere
  20-odd years of not having ever encountered round-to-even.
* I understand why round-to-even makes sense. I also understand that
  it's the IEEE-preferred way of doing things, and that it's used by
  Java. (no, I'm not really a Java developer, I just play one as part
  of my sysadmin role)
* Emacs is inconsistent across platforms in its operation, which makes
  the function a bit less useful than it might be. I believe that
  Emacs should be consistent across platforms, and if the underlying
  OS is 'broken' then the Emacs implementation should correct for it.
* The function is underdocumented. If there are two different ways of
  doing a thing, and they produce different results, then the one
  chosen in the implementation should be explicitly mentioned. Well,
  in this case, anyway.

All IMHO, as appropriate.

Cheers,
Waider.