Re: [Json] The names within an object SHOULD be unique.

[Matthew Morley <[e-mail]>]

On Wed, Jun 5, 2013 at 11:44 PM, Joe Hildebrand (jhildebr) <
[e-mail]> wrote:

> On 6/5/13 7:56 PM, "Matthew Morley" <[e-mail]> wrote:
>
> >The best way to handle 'comments' is to have them be first class citizens
> >in the data. It is a topic that goes back a long way, and since they were
> >removed with intention then, using ambiguity in the specification to
> >re-add them seems like a poor approach
> > now.
>
> Do you have a backward-compatible proposal?  I can't think of one, but
> that doesn't make it impossible.
>

Nope :)

I was meaning that the comment should be in data structure itself, existing
as a pattern of usage in JSON, rather than part of the JSON specification.


There are examples of adding "_comment_{foo}":"comment on pair foo" to
objects and insertion of {"_comment_0":"comment on index 0"} into arrays.
With the expectation that those comment keys and comment objects be
stripped if their existence will break schema expectations. But it goes a
bit further than just that, in that comments on a top level array or object
would require the payload being wrapped in an array with the 0 index being
the comment object, and the 1 index being the real payload.

It is the unordered nature of keys, that makes the use of duplicate keys
for comments seem like a very strange choice. Every time I look at a JSON
object, I try to imagine the keys (usually on a single line with the value)
constantly shifting their order and cycling around. That is the reality of
what can happen as it encodes/decodes and gets passed around.

-- 
Matthew P. C. Morley