Re: [Json] Counterproposal #2 on work items

[Tim Bray <[e-mail]>]

Which was why I was interested in a best-practices spec.  Because it's
basically crazy, in a network protocol, to issue duplicate keys, whether or
not it's legal JSON.  Somewhere this should be stated.

On Feb 20, 2013 9:15 PM, "Robert Sayre" <[e-mail]> wrote:
>
>
>
> On Wednesday, February 20, 2013, Cyrus Daboo wrote:
>>
>> Hi Vinny,
>>
>> --On February 20, 2013 8:02:13 PM -0800 Vinny A <[e-mail]>
wrote:
>>
>>> I believe org.json throws a exception if it hits a duplicated key; at
>>> least the documentation [
>>>
http://json.org/javadoc/org/json/JSONObject.html#JSONObject(java.lang.Str
>>> ing) ] suggests that to be the case. I don't remember where I saw the
>>> append to array parser - but it's such a good idea it's worth bringing
up
>>> in this conversation.
>>
>>
>> Another data point: Python 2.7.2 -
>>
>>>>> import json
>>>>> print json.loads('{"a":"a", "a":"b"}')
>>
>> {u'a': u'b'}
>>
>> So Python will accept duplicate keys, but one value is lost. This makes
sense since keys in Python dicts are unique. At this point I favor changing
the SHOULD to a MUST in RFC 4627bis to make it clear that there is lack of
interoperability here.
>
>
> There is a lack of interoperability, and this issue is one of the smaller
points. The larger point is that the JS implementations (and many smaller
ones) have no incentive to change their behavior, and what they do is
reasonable. Let's just do that.
>
> - Rob