Re: [Json] I-JSON Tpic #2: Top-Level

[Nico Williams <[e-mail]>]

On Mon, Apr 28, 2014 at 4:16 PM, Tim Bray <[e-mail]> wrote:
> RFC 7159 already says that for interoperability, top level objects should be
> objects or arrays.  At the simplest level, if all I-JSON does is forbid all
> the things that 7159 calls out as interoperability problems, that would be
> good enough for me.  The notion that i-json would *weaken* 7159’s
> requirements by allowing a top level object to consist of just true or
> "true" seems outlandish to me.

That was a hard-won but-grudging concession to reality.  All encoders
and parser I checked have long had an option to permit scalars at the
top-level; RFC7159 should have required that by default scalars be
accepted at the top-level given that such things are found in the
wild.

I don't really care to re-argue that -- grudging or not, the
concession was all that mattered.  But if we're going to argue about
what I-JSON should do about the top-level based on the fact that the
concession in RFC7159 was grudging, then I guess we'll have to
re-argue this!

But I'd rather focus on why one might want to mandate objects at the
top-level, what to do when one has an array (use text sequences?!),
and what to do when this isn't appropriate ("don't use I-JSON" being
fine, provided we don't end up squeezing out non-I JSON).

Nico
--