Re: [Json] JSON Concluded? Well, maybe not

[Anders Rundgren <[e-mail]>]

On 2018-01-01 00:39, Richard Gibson wrote:
> On Sun, Dec 31, 2017 at 4:34 PM, Anders Rundgren <[e-mail] <mailto:[e-mail]>> wrote:
> 
>     For normalization it turns out that JSON.parse() and JSON.stringify() as specified by ECMA and already supported by the most widely available JSON tools is all you need!
> 
> 
> That is definitely not true. A trivial counterexample: https://jsbin.com/luyoveyuqu/edit?js,console,output

Right, JSON.stringify() doesn't not canonicalize but preserve property order [1] + normalize data which for "crypto-safe" JSON applications is entirely sufficient.

If you go back to your example you can see another nicety of the EMCA parse/stringify scheme: JSON data is actually generated in the order specified by the programmer.

Or as I like to put it: JSON grew out of JavaScript, now it has come home again :-)

Cheers,
Anders

1] Extract from an I-D in preparation:

For JavaScript optimization reasons, ECMAScript's JSON.parse()
internally rearranges order of properties with names expressed
as integers, making a parsed JSON string like
'{"2":"First","A":"Next","1":"Last"}' actually serialize as
'{"1":"Last","2":"First","A":"Next"}'.  Due to this fact,
signature creators must (in the hopefully rather unusual case
cross platform applications would mandate numeric property names),
in an here unspecified way, "emulate" this scheme since this
behavior is not intended to be an additional requirement to
support by JSON tools in general in order to use this specification.