Re: [Json] Proposal: the minimal edit

[Norbert Lindenberg <[e-mail]>]

On Jun 22, 2013, at 11:52 , John Cowan wrote:

> R S scripsit:
> 
>> It's my opinion that none of the errors or inconsistencies present in
>> the document are significant, since JSON works pretty well.
> 
> It does at the level of an informational protocol, but not at the level of
> standards track.  Would you be willing to sign a contract, arm's-length,
> with someone whose only constraint was that their code passed *some*
> interpretation of the RFC?  I certainly would not.  As I pointed out
> before, the parser conformance clause (first paragraph of Section 4)
> is satisfied by any program that accepts arbitrary input and produces
> arbitrary output distinct from the input (or indeed no output at all,
> given the over-free wording in the second paragraph).


I agree with John - for a standard, some of the errors, undefined terminology, and inconsistencies seem unacceptable. The ones that worry me in particular (I tend to worry about internationalization):

- The RFC uses the term "Unicode character", which isn't defined in the Unicode standard as far as I can tell. If it means "Unicode assigned character", the most likely candidate, then it's inconsistent with the "unescaped" production in the grammar, which allows all Unicode code points.

- The RFC says "JSON text SHALL be encoded in Unicode", which is obviously meant as a normative statement, but, as John pointed out, is meaningless because Unicode is not an encoding. Other text in the RFC seems to assume that "encoded in Unicode" means "encoded in UTF-8, UTF-16BE, UTF-16LE, UTF-32BE, or UTF-32-LE".

Norbert