Re: [Json] Wording on encoding; removing the table

["Martin J. Dürst" <[e-mail]>]

I agree with Björn here. Removing UTF-16 and UTF-32 from the spec 
completely opens the door for any other odd encoding, which would be a 
net decrease in interoperability.

Regards,   Martin.

On 2013/11/23 8:02, Bjoern Hoehrmann wrote:
> * Paul Hoffman wrote:
>> Proposed replacement:
>>
>>    The default encoding for JSON transmitted over the Internet is UTF-8.
>>    Transmitting JSON using other encodings may not be interoperable
>>    unless the receiving system definitively knows the encoding.
>>
>> Does anyone have a technical objection to the proposed replacement? If
>> so, please state the error and (hopefully) a correction.
>
> It is necessary for reasons of security and interoperability that all
> application/json processors agree on how to get from the sequence of
> bytes that make up the application/json entity to a sequence of integers
> that are used in the ABNF definition of the JSON syntax. For example,
>
>    data:application/json,%5B%22Bj+APY-rn%22%5D
>
> Under the rules you propose, this can be interpreted as if it were
>
>    ["Björn"]
>
> An implementation that interprets it thus is fully conforming because
> the bytes look like they are UTF-7 encoded text and the specification
> does not unambiguously say that, no, the implementation must not take
> it as UTF-7 encoded document, instead it must use UTF-8 to decode.
>
> Under the rules of RFC 4627 there can be no UTF-7 encoded application/
> json entities at all and processors are required to decode the example
> using UTF-8.
>
> Some web browsers had exactly this problem for a long time with other
> formats, it is entirely unacceptable to go back to "just do whatever"
> specifications for character encoding determination.