Re: [apps-discuss] feedback on draft-ietf-appsawg-json-patch-01.txt

[Julian Reschke <[e-mail]>]

On 2012-03-21 07:55, "Martin J. Dürst" wrote:
> On 2012/03/21 6:40, Murray S. Kucherawy wrote:
>
>> Does there need to be one-and-only-one?
>
> In theory, not necessarily. In practice, not having a lot of variation
> usually helps.
>
>> Or I guess more accurately, do we need to say this is the one and only
>> way to do this? Seems like that closes the door to someone coming up
>> with a better idea.
>
> Fragment pointers into datastructures such as JSON isn't exactly a brand
> new field with lots of new possibilities. And XPointer has shown that
> making it too complicated cuts out most implementations. My proposal
> would be to somehow include an extensibility hatch, but then say this is
> the one and only one.

In XML, as far as I understand, the escape hatch is that identifiers do 
not allow brackets, and thus these are free for defining new addressing 
schemes. So in JSON Pointer, we'd need to escape a few more characters 
when using in fragment identifiers.

Also, it seems this needs to be coordinated with a potential spec 
defining a "+json" media type suffix...

Best regards, Julian