Re: [Json] -0.0

Hmm.   Many other languages, for example C, are not adequately specific
concerning the interpretation of floating point constants, particularly -0.0

However, I don't know of any language or formalism that is used in a way
that is so cavalier about the interpretation of numeric constants in
general, so I would have to say that the breadth of the problems with -0
are specific to JSON.  In all the other languages and formalisms that I
have used -0 is integer 0, and thus the same as 0, and this has been
explicitly or implicitly stated in the definition of the language or
formalism.  Even C with one's complement integers should get this right,
although direct casting to unsigned integers might expose the existence of
signed zeros.

Peter F. Patel-Schneider
Nuance Communications

PS:  Yes, yes, I expect that there are *many* languages and formalisms that
are just as bad as JSON with respect to numbers.  I just haven't had the
displeasure of needing to work with any of them yet.

On Sun, Sep 29, 2013 at 1:46 PM, Tim Bray <tbray@textuality.com> wrote:

> Yeah, but my opinion is shifting here... it's starting to feel to me like
> there's nothing remotely JSON-specific about the problems around -0.  Let
> me see, if I were going to do another draft based on what people have been
> saying here, I'd, in Section 6
>
> - Remove "are represented without "frac" or "exp" parts (for example as 3
> not 3.0)"
> - Lose the whole paragraph beginning "Numbers which represent zero without
> a sign...
>
> Less is more. Is anyone passionate about keeping either?
>
>  -T
>
>
> On Sun, Sep 29, 2013 at 1:33 PM, Carsten Bormann <cabo@tzi.org> wrote:
>
>> On Sep 29, 2013, at 22:25, R S <sayrer@gmail.com> wrote:
>>
>> > I am not sure that file is relevant
>>
>> It is an example of an implementation that encodes a floating point
>> negative zero as »-0«, which is then decoded as a plain (integer) zero by
>> most implementations.
>>
>> Summary: If you want wide interoperability, don't rely on negative zero
>> (in any form) staying distinct.  (And if you don't need wide
>> interoperability, the encoded form »-0.0« appears to survive best.)
>>
>> Grüße, Carsten
>>
>> _______________________________________________
>> json mailing list
>> json@ietf.org
>> https://www.ietf.org/mailman/listinfo/json
>>
>
>
> _______________________________________________
> json mailing list
> json@ietf.org
> https://www.ietf.org/mailman/listinfo/json
>
>