Re: [apps-discuss] JSON Patch

[Tony Hansen <[e-mail]>]

On 11/23/2011 1:06 PM, Julian Reschke wrote:
> On 2011-11-23 19:02, Paul C. Bryan wrote:
>> On Wed, 2011-11-23 at 18:37 +0100, Julian Reschke wrote:
>>> On 2011-11-23 18:19, Paul C. Bryan wrote:
>>> >  I'm now inclined to simply make"to"  a JSON pointer as well.
>>> >  Semantically it's clear, It's easy to apply, and JSON diff
>>> >  implementations can be smart if they want and notice something moved
>>> >  from one node in the graph to another.
>>>
>>> +1, except that I'm not convinced that leaving this optional is a good
>>> idea...
>>
>> I'd say if JSON pointer is adopted for "to", it should only be JSON 
>> pointer.
>>
>> Paul
>
> Works for me.
>
> (What I had in mind was to consider something not starting with / as 
> relative pointer...)

If you introduce relative pointers, you also need the concept of .. to 
move up the path, and have to define semantics for moving past the 
beginning of the path, and the concept of normalizing a path, and 
probably a dozen other things that are normally part of relative pathing.

In thinking about the example, I find "/foo/2" much clearer than just 
"2" as the target.

Given that Javascript uses 0-based arrays, should we also be using 
0-based paths here? Or has that horse already left the barn?

     Tony Hansen