Re: [Json] [apps-discuss] merge-patch in APPSA and i-json in JSON

[Paul Hoffman <[e-mail]>]

On Jul 8, 2014, at 12:06 PM, John Cowan <[e-mail]> wrote:

> Paul Hoffman scripsit:
> 
>>> - Is merge operation defined for target JSON that isn't i-json?
>> 
>> Yes. The merge operation is defined for all targets, regardless of
>> the targets' profile of JSON.
> 
> Actually, it isn't.  It is far from clear from
> draft-ietf-appsawg-json-merge-patch-04 what to do if a key appearing in
> the patch matches more than one key appearing in the target.  This has
> nothing to do with I-JSON as such, but is a deficiency in the merge-patch
> algorithm, the relevant part of which I reproduce here:
> 
>       for each Key/Value pair in Patch:
>         if Value is null:
>           if Key exists in Target:
>             remove the Key/Value pair from Target
>         else:
>           Target[Key] = MergePatch(Target[Key], Value)
> 
> What is meant by "the Key/Value pair" if there is more than one?
> Are all such pairs to be removed or mutated, or only one of them?

The horse's carcass is so decomposed that you can't even find what to kick anymore. RFC 7159 says "When the names within an object are not unique, the behavior of software that receives such an object is unpredictable." It would be actively harmful for draft-ietf-appsawg-json-merge-patch to say "we can make it predictable in our own unique fashion".

--Paul Hoffman