Re: An additional thought Re: Some thoughts on Motion 19 (still under vote)
> Date: Thu, 16 Sep 2010 05:14:54 -0500
> From: Ralph Baker Kearfott <rbk@xxxxxxxxxxxx>
> To: rbk@xxxxxxxxxxxxx
> CC: Paul Zimmermann <Paul.Zimmermann@xxxxxxxx>, stds-1788@xxxxxxxxxxxxxxxxx
> Subject: An additional thought Re: Some thoughts on Motion 19 (still under vote)
>
> Paul, P-1788,
>
> Yes, and I indeed had known that (but perhaps forgotten).
> I'm not sure about staggered arithmetic, though, and it's
> becoming more visible. (There also had been a thread
> discussing variable precision.)
>
> Nonetheless, the most common, in my perception, view of 754
> is the plain 64-bit binary arithmetic. Furthermore, the
> "keep to 754" motion was withdrawn, with more "no" votes
> than "yes." Thus, previous actions of this working group
> do not fetter us with regard to 754. A possible final 1788
> standard might specify two 754 64-bit binaries to represent
> intervals. That would lead to a particularly simple standard,
> but would not give even minimal specifications for other
> arithmetics. Another possibility would be to have interval
> types corresponding to each 754 data type. All of this has
> been discussed in relation to 1788, if not on this mailing
> list, by smaller groups. An even more complex standard would
> involve all of the 754 data types, and additionally explicitly
> specify both inf-sup and mid-rad.
>
> I think motion 19 was meant to be a compromise to maintain
> some level of simplicity while providing minimal specifications
> to other arithmetics that may be common. The committee needs
> to decide whether or not we need to do that and whether or not
> motion 19 accomplishes that.
>
> Baker
>
All this is so. And, as you suggest, none of it
prevents us from singling out intervals that use 754
arithmetic as a distinguished class of interval
datatypes with their own, more stringent, rules.
John suggests as much in motion 19 section 3.4.
Although I believe John kept it as just a suggestion
so as not to further complicate the motion with what
is really a side issue.
And, should we consider distinguished classes of 754
intervals in the future, they can apply to arbitrary
precision binary & decimal arithmetic as well.
As Paul correctly points out.
One step at a time...
Dan