Re: [Json] Call for Consensus: Proposed Text for "8.1 Character Encoding"

[Pete Cordell <[e-mail]>]

On 18/04/2017 06:22, Martin J. Dürst wrote:
> On 2017/04/18 05:47, Carsten Bormann wrote:
>> On Apr 17, 2017, at 19:56, Nico Williams <[e-mail]> wrote:
>>>
>>>> Thinking about this more, putting an encoding detection algorithm as an
>>>> appendix seems like a reasonable compromise to me.  To start, how about
>>>> removing the detection text from Section 8.1 and have an appendix that
>>>> starts with that text plus the table?
>>>
>>> Or we could even just assert that such an algorithm is possible, and
>>> that implementors MAY implement one.
>>
>> Indeed.
>>
>> Broken record mode:
>>
>> — writing up the algorithm sounds like encouraging implementation.
>>   We *don’t* want people to implement this!
>>   (The whole interminable non-UTF-8 saga probably just was a nod from
>> the RFC 4627 authors to the remnants of UTF-16 land, which mostly have
>> died off since.  Why resurrect?)
>>
>> - there have been about 15 attempts to define this algorithm on the
>> mailing list.
>>   All were wrong.
>>   An Internet Standard should contain tried and true material, not
>> errata fodder.
>>
>> - an implementer is in a much better position to get this right than
>> the standard, because they can write unit tests.
>
> I completely agree with Carsten. As far as I know, and as far as we have
> been told on this list, if some JSON isn't in UTF-8, then it simply will
> not interoperate.

+1

If we do do this, I think we could add some example test messages to 
helps with the development, e.g.:

     {"Example":1}
     {}
     "Example"
     ""
     "U+0100" (where U+0100 is the UTF form of the character, not ASCII)
     1

> In my view, the only reason to still have a MAY for UTF-16/32 is that
> this will avoid questions like: "I have a JSON parser in language FOO,
> it can take a string or an input stream as an argument. In FOO, strings
> are UTF-16, but the JSON RFC doesn't seem to allow this. What should I do."

IMO the answer to that is, "that's why it says 'JSON text _SHOULD_ be 
encoded in UTF-8'".

I agree with John Cowen, that use of UTF-16/32 is purely for internal 
scenarios; not on the Internet.  As such, I believe the IETF is going 
beyond its remit to say you can use UTF-16/32 for your internal purposes 
that I know nothing about and care nothing about, but you're not allowed 
to use other encodings that maybe more natural for your system.

Pete Cordell
Codalogic Ltd
C++ tools for C++ programmers, http://codalogic.com
Read & write XML in C++, http://www.xml2cpp.com