Re: Why (IMO) you should vote Yes to Motion 14.02
> Date: Fri, 02 Jul 2010 17:03:00 -0700
> From: "G. William (Bill) Walster" <bill@xxxxxxxxxxx>
> To: "Corliss, George" <george.corliss@xxxxxxxxxxxxx>
> CC: Dan Zuras Intervals <intervals08@xxxxxxxxxxxxxx>,
> P1788 <stds-1788@xxxxxxxxxxxxxxxxx>,
> Nate Hayes <nh@xxxxxxxxxxxxxxxxx>
> Subject: Re: Why (IMO) you should vote Yes to Motion 14.02
>
>
>
> Dan,
>
> I believe you are headed in the right direction.
>
> One suggestion: Why not leave "tightness" of conversions from level 1 to
> level 2 and back be quality of implementation features? Why can't the
> only requirement be to contain the level one set of results obtained by
> performing the given operation or expression evaluation on the tightest
> possible conversions of level 2 operands to level 1?
>
> Cheers,
>
> Bill
>
> . . .
>
Bill,
I think we're going in a different direction since
I made those remarks but I will try to answer your
question as asked anyway.
First, we are trying to standardize on something.
Just what we are trying to standardize is still up
for grabs. But, at the very least, I think that
means that we are trying to get some sort of
predictability from one implementation to the next.
Second, there is the fact that requiring that an
implementation shall return the tightest possible
containing interval is so easy to do these days.
At least for the basic operations in isolation.
Larger expressions & transcendental functions are,
like the standard itself, still up for grabs.
And, third, for good practical & political reasons,
we may have to give up on tight specifications
anyway. So you may be correct that it will turn
out to be a quality issue.
But I would rather not standardize implementations
of poor quality.
Time will tell.
Dan