Re: [Json] Limitations on number size?

[Tatu Saloranta <[e-mail]>]

On Tue, Jul 9, 2013 at 8:38 AM, Nico Williams <[e-mail]> wrote:

> On Tue, Jul 9, 2013 at 10:27 AM, Tim Bray <[e-mail]> wrote:
> > Yeah, we can’t retroactively change the 4627 production for number, nor,
> in
> > the -bis, say very much useful aside from a caution that implementations
> are
> > all over the map.  When we move on to the BCP or Internet JSON or
> whatever,
> > I think the single most interesting discussion will be how to constrain
> > numeric values for maximum interoperability.
>
> I would like to do two things here though:
>
> 1) give advice as to what numeric ranges and precisions are generally
> supported (e.g., 32-bit signed integers);
>
> 2) what a parser SHOULD do when it parsers a number that it cannot
> represent.
>

This would be useful. One challenge is that of precision for floating-point
types: figuring out whether loss occurs is rather costly. For integer types
this is easier.

-+ Tatu +-