Re: [apps-discuss] Concise Binary Object Representation (CBOR)

[Phillip Hallam-Baker <[e-mail]>]

On Thu, May 23, 2013 at 3:55 PM, Carsten Bormann <[e-mail]> wrote:

> On May 23, 2013, at 21:31, Phillip Hallam-Baker <[e-mail]> wrote:
>
> > Completely off topic here but raised by the discussion. JSON needs to
> address NaN encoding.
> >
> > NaN is essential for exchange of numeric data in scientific
> applications. JSON is going to be used in those applications. It would be
> better to have one way to use JSON in scientific applications rather than
> have different work arounds emerge.
>
> The applications I have seen trying to address the problem have simply
> used null to encode NaN.
> There is rarely a case when you need to be able to express both null and
> NaN in one place (use CBOR, then :-).
>

There are many work-arounds possible. A standard should only have one.



> A bigger problem is -Infinity/Infinity, but similar hacks can be used
> there, too (say, use false/true, or stringify).
>
> > There is an ongoing effort to document JSON, this should be included in
> those efforts.
>
> JSON MUST NOT change, so I don't think I agree with that (neither does the
> proposed charter of the WG).
>

I don't think we have ever accepted a standard that declares it can't be
changed. We have certainly changed all the ones we started with. Even ASCII
was changed.

There are several ways that NaN and +/- infinity can be introduced without
changing JSON syntax. The point is there should only be one way that code
does so.


-- 
Website: http://hallambaker.com/