Re: [Json] Limitations on number size?

[Jacob Davies <[e-mail]>]

On Tue, Jul 9, 2013 at 3:58 PM, Peter F. Patel-Schneider
<[e-mail]> wrote:
> The only suitable guidance provided in RFC4627 is via ECMAScript and
> ECMAScript is firmly IEEE floating point double only.
>
> So why are you surprised that I came up with this conclusion?

What I found surprising was that there appeared to be lots of people
willing to say that the strings in the JSON data model (not the
format) are Unicode, or that the keys in the objects must be unique
according to one particular method of string normalization, but not
that the numbers in JSON are derived from Javascript and that
interoperability in the most important use of JSON (sending data to
browsers) absolutely requires you to stay within the size limit of
Javascript numbers (and failure to do so has led to some notable bugs
that require backwards-incompatible API changes, i.e. going from
number -> string for some field).

My take prior to this discussion was that the mild annoyance of not
being able to send 9223372036854775807 as a number in JSON was the
reasonable price you paid for having an interoperable number range in
the format. It seemed like most people here disagreed with my
impression, which is fair enough, and it sounded like Douglas
Crockford's intent was not to restrict them in such a way, but I would
be surprised if many people actually - intentionally at least! - use
numbers from outside the Javascript range in JSON, given the problems
it causes for Javascript.

As for numbers smaller than that range, the RFC says implementations
may reject numbers they don't like, so an implementation without
floating point numbers is OK, just not capable of general
interoperability.