Re: [Json] Comments on proposed charter for JSON

[Paul Hoffman <[e-mail]>]

On Mar 4, 2013, at 11:57 AM, Bjoern Hoehrmann <[e-mail]> wrote:

> * Paul Hoffman wrote:
>> On Mar 4, 2013, at 11:12 AM, Bjoern Hoehrmann <[e-mail]> wrote:
>>> * Paul Hoffman wrote:
>>>> I *do* want to see browser folks familiar with the ECMAscript spec to 
>>>> start working on a specific list of differences between it and RFC 4627. 
>>>> If y'all can do that before the BoF meeting next week, it will help the 
>>>> charter discussion.
>>> 
>>> That's already been done in the ecmascript specification:
>> 
>> Incompletely. Have you already forgotten about:
>> 
>> NOTE In the case where there are duplicate name Strings within an 
>> object, lexically preceding values for the same key shall be 
>> overwritten.
> 
> I do not regard that as a difference beyond that ecmascript defines an
> API while RFC 4627 a data format. What would be the change here that
> could be adopted? Require implementations to ignore lexically preceding
> values, i.e., they must not be reported to higher-level applications and
> implementations must not treat duplicates as an error of any kind?

RFC 4627 is not purely a data format: there are implied parsing rules. For all the parts of 4627 format that are MUST-level, the implied parsing rules are clear for developers; where there are SHOULD-level rules in the format, the parsing rules become indeterminate.

It would make sense for 4627bis to deal with parsing rules in a way that matches ECMAscript (minus the optional reviver). It can do this by making everything MUST-level, or it can have a short set of parsing rules. I *thin* that the only change that is needed would be to add the "NOTE" above because that is the only "SHOULD" in 4627, but I want to hear from people who have spent much more time with ECMAscript to be sure.

--Paul Hoffman