Re: [Json] Proposal: the minimal edit

[Paul Hoffman <[e-mail]>]

<no hat>

On Jun 24, 2013, at 6:35 PM, Norbert Lindenberg <[e-mail]> wrote:

> 
> On Jun 24, 2013, at 17:28 , Paul Hoffman wrote:
> 
>>>> - ECMAScript implementations can generate and consume code points in JSON strings that are not
>>>> Unicode characters.
>>> 
>>> Which definition of "Unicode character" are the three of you referring to?
>> 
>> Does it actually matter? This thread is about making minimal changes to the current spec. The current spec says "Unicode characters". The proposed addition to the spec indicates that, for some definition of "Unicode character" that is not in RFC 4627, ECMAScript can have strings with more than just Unicode characters.
>> 
>> Do you believe that proposed sentence is incorrect?
> 
> Without a definition of "Unicode character" we can't determine whether the sentence is correct or not.

I believe you can. If you look at the mailing list, you will see that some people interpret ECMAScript as allowing every code point in strings. There are some code points that do not represent Unicode characters. Isn't that sufficient?

> On the other hand, all code points that an ECMAScript implementation can generate and consume are allowed by the JSON grammar, so by that measure there's no difference between the two specifications.

There is wide disagreement on that statement in the WG so far.

> I think what Tim and others really want to say is "JSON over the wire must not use surrogate code points", with Tim adding "... or non-character code points". From that would follow that "ECMAScript implementations can generate and consume code points in JSON strings that must not be used in JSON over the wire". And I think that's the kind of information that actually matters from an interoperability point of view. It's not a minimal edit though.

Exactly right. The attempt in this thread is *not* to list interoperability points, just cleanup.

--Paul Hoffman