Re: [Json] Update Standard to support ECMA-262 BigInt

[Rob Sayre <[e-mail]>]

On Thu, Apr 29, 2021 at 4:38 PM Tim Bray <[e-mail]> wrote:

> Yeah, speaking as the editor of a bunch of those RFCs, I would very much
> like it if JSON could be extended to have comments, timestamp literals, and
> typed numbers.
>

Well, if JSON had comments, it would have all sorts of extensions by now,
which is why it doesn't have them. Comments that are actually comments
would be nice, but programmers can't be trusted.


> But it's too late. JSON is finished, is generated and consumed by a very
> large amount of deployed software which is simply not going to be updated;
> changing the syntax at this point would be catastrophic.
>

Yes.


> For crypto apps, you pretty well have to put the big numbers in strings.
>

JSON numbers interchanged between languages that have arbitrary-precision
numbers work fine. But, JS BigInt (original post) operations aren't
constant time, so you shouldn't use them for crypto anyway [1].

Put the big numbers in strings if you think JavaScript might be in the
picture. For many other languages, you don't need to. It should also be
possible to detect number precision in a JSON encoder/decoder if really
desired.

thanks,
Rob

[1]
https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/BigInt#cryptography
(see also the suggestion for JSON stringifying below)