Re: [apps-discuss] Confusing JSON Pointer

["Paul C. Bryan" <[e-mail]>]

On Wed, 2012-03-28 at 08:47 +0900, 지송원 James Songwon Chi wrote:

> 2012년 3월 27일 오후 11:23, Julian Reschke <[e-mail]>님의 말:
> > On 2012-03-27 15:49, 지송원 James Songwon Chi wrote:
> >> 2012년 3월 27일 오후 4:57, Julian Reschke<[e-mail]>님의 말:
> >>> On 2012-03-27 07:22, 지송원 James Songwon Chi wrote:
> >>>>
> >>>> After reading JSON Pointer (draft-ietf-appsawg-json-pointer-01), I am
> >>>> confused with encoding or escape scheme.
> >>>> For me, it is an ambiguous scheme. For example, a reference token
> >>>> "ref/1" can be expressed following ways:
> >>>>
> >>>> 1) If it is a URI fragment identifier and a JSON Pointer
> >>>>       Step 1. apply URI encoding: ref%5C1
> >>>>       Step 2. apply JSON Pointer encoding:  ref%5C1
> >>>>       Final result: ref%5C1
> >>>>
> >>>> 2) If it is a JSON Pointer and a URI fragment identifier
> >>>>       Step 1. apply JSON Pointer encoding: ref^/1
> >>>>       Step 2. apply URI encoding:  ref^%5C1
> >>>>       Final result: ref^%5C1
> >>>>
> >>>> I expected the same results.
> >>>>
> >>>> And if my interpretation is wrong, to prevent the similar question,
> >>>> there should be some examples for this kind.
> >>>> ...
> >>>
> >>>
> >>> I don't understand the question.
> >>>
> >>> The spec states how to put JSON Pointers (^-escaping) into fragment
> >>> identifiers (%-escaping).
> >>>
> >>> As far as I can tell the outcome then is what you have under 2), no?
> >>>
> >>> Best regards, Julian
> >>
> >> Thank you Julian for the reply.
> >> Yes. Answer is 2) for above question. Well, but I gave a wrong
> >> question and a wrong example (And solidus is 0x2F, not 0x5C.)
> >>
> >> This is corrected one:
> >>
> >> If I want a JSON Pointer to be represented as both a JSON String and a
> >> URI fragment identifier, I should apply three encoding scheme.
> >> There are two sequences with example:
> >> ...
> >
> > You convert *either* to a string *or* to a URI fragment. Never both.
> I don't agree with you. First, there is no such words.
> Second, if it is used in a JSON Schema as a "$ref" value, it should be
> both JSON String and URI.


The intent of the draft is to establish rules for encoding in a JSON
string or in a URI. Of course, I agree that a URI can in turn be
expressed in a JSON string, but in this case, URI encoding applies; you
do not mix-and-match encodings.


> By the way, do we really *need another escape* even there are two
> encodings already?


I'm not sure what you mean here. If you encode as URI, presumably all
valid URI characters are representable without any further escaping in a
JSON string.

Paul