James Giles said:
> Malcolm Cohen wrote:
> ...
> >> I was told that my request was removed from the list of feature
> >> requests and placed on a list of interp requests. Well, there is
> >> no way that an interp can satisfy the current common practice
> >> without completely contradicting the present wording of the
> >> document.
> >
> > Piffle and nonsense. Your interpretation is does not make sense,
> > because the standard would thereby be inconsistent (contain a
> > contradiction). The fact that practically NO-ONE ELSE agreed with you
> > that that was the intent of the standard might be a clue here.
>
> So, the actual words:
>
> Processors that distinguish between positive and negative zeros
> shall treat them as equivalent
Right, and is that computationally equivalent? mathematically equivalent?
Physically equivalent?
> ...
> (2) as actual arguments to intrinsic procedures other than those for
> which it is explicitly specified that negative zero is distinguished,
>
> Actually means the opposite of what it says? Double speak.
> That's why I called the practice reprehensible.
Balderdash. The authors of the words have said that they wanted
zero to be treated *as zero*, whether it is negative or positive.
Ancient History: some machines with negative zero actually produced
bad results (i.e. they were not treated as zero) in some situations.
That is, according to the authors of the words, what those words are
trying to prohibit.
> (Nor is there any contradiction between my position and any
> part of the standard. Point to anything actually *in* the document
> that requires intrinsic functions to actually distinguish the sign
> of zero.)
The TRANSFER intrinsic.
Direct contradiction - either the result has the physical representation
of positive zero or negative zero, it cannot have both.
> > And even if it did, so what? Your argument is completely irrevelant:
> > the process is about fixing defects in the standard.
>
> Then fix them in the standard, not in an obscure document
> no one reads.
They *ARE* fixed in the standard! If you get a new copy of the standard
from ISO it comes with the corrigenda attached.
> ...
> > When the standard contains errors, it is the DUTY of the working group
> > to fix them. [...]
>
> Agreed. Fix them *in* *the* *document*. Don't change the
And THAT IS EXACTLY WHAT IS DONE.
> meaning in an interp and then *claim* that's actually what the
> document says.
No, THAT IS NOT WHAT THE CLAIM IS.
Really, one would think you've never read any of these documents.
> If the words of the document change, it's not an interp but
> a change. If the words don't change, and the interp contradicts
> the document, one of them is an exercise is dishonesty.
Why are you harping on about this "interp" word?
THAT IS WILFUL MISCHARACTERISATION.
The USA process calls the queries "Requests for Interpretation", but the
result of a request is only an "Interpretation" if the words don't change.
If the words change it is called an Erratum. Looking at ANY of the
actual responses in the last 15 years would have revealed that.
The ISO process (which is the one which actually changes the standard
by producing corrigenda) is Defect Processing. Nothing about "interp".
Cheers,
--
........................Malcolm Cohen ([log in to unmask]), Nihon NAG, Tokyo.
|