Re: [Json] Call for WG Consensus to Adopt I-JSON as a WG Item in the Charter

[Phillip Hallam-Baker <[e-mail]>]

On Mon, Mar 17, 2014 at 11:21 AM, Nico Williams <[e-mail]>wrote:

> On Mon, Mar 17, 2014 at 6:56 AM, Phillip Hallam-Baker <[e-mail]>
> wrote:
> > On Sun, Mar 16, 2014 at 11:51 PM, Nico Williams <[e-mail]>
> > wrote:
> >> My main complaint is with the recommendation/requirement that JSON
> >> texts be objects at the top-level.
> >>
> >> Instead I would say that if there's reason to need to signal "schema"
> >> metadata in-band then a) application should specify how they do that,
> >> b) here's a few ways in which they might, and all their pros and cons.
> >>
> >> Note too that there are many apps where there's no need to signal
> >> schema metadata in-band because they always have some way to
> >> communicate relevant metadata out of band.
> >
> >
> > I have no problem with this as a convention in a protocol. But I don't
> see
> > the need for that restriction in I-JSON.
>
> If the purpose is to signal schema metadata then I'd like to explore
> other conventions and go for a "you must provide such metadata"
> statement in order to make yours an I-JSON usage.
>
> > In a protocol design the spec is going to say something like 'the
> encoding
> > is in the I-JSON subset of JSON'. Then its going to describe the format
> of
> > the data which for a protocol is almost always gonna be an object per
> > message at the top level. Though it could be an array of objects if
> multiple
> > requests per message are supported.
>
> Right, imagine wrapping an array of objects in an object just to
> denote schema metadata.  What a pain.  You're probably streaming the
> array, so now you gotta add an object around it -- and remove it -also
> in a streaming manner- on the parser side.
>
> Or perhaps you have a sequence of objects to return... should you mark
> up every one of those objects?  While streaming them?  More pain.
>
> We should consider how this markup happens.
>
> Incidentally, we could say that the first text in a JSON text sequence
> should denote schema metadata for the rest...  That would not be so
> painful.  (although it would still interfere with existing usage).
>
> Nico
> --
>


As a general rule, whenever I see a proposal that depends on the order of
JSON objects, I think that either an Array or an Object needs to be
introduced.

I very much prefer this

{ "get-request" : {  .... request data here .... } }


to this:

{ "jscommand" : "get-request",
   ... rest of request data follows ...
   }


One of the things I have come to detest in HTTP is that it does not
separate protocol headers from content metadata. There really should be a
separation between the two so that it is possible to do things like signing
the content metadata.


-- 
Website: http://hallambaker.com/