Re: [Json] Proposed minimal change for duplicate names in objects

[Tim Bray <[e-mail]>]

There’s no need to mention streaming; to start with, we’d have to define a
bunch of terms.

Say, as the RFC does, that you SHOULD NOT emit duplicates because it leads
to variable results, and also note that there is software in the world that
nonetheless does, possibly note that ECMA allows it, and say that receiving
software SHOULD do last-dupe-wins.  JSON is what JSON is.

For “Internet JSON”, it’s a different argument. -T


On Wed, Jul 3, 2013 at 2:39 PM, Nico Williams <[e-mail]> wrote:

> On Wed, Jul 3, 2013 at 3:44 PM, Tim Bray <[e-mail]> wrote:
> > On Wed, Jul 3, 2013 at 1:11 PM, John Cowan <[e-mail]>
> wrote:
> >>
> >> Pete Resnick scripsit:
> >>
> >> > So I don't think saying RECOMMENDED or SHOULD regarding not
> >> > producing duplicates should be seen as a non-starter. That's a
> >> > defensible position for the WG to take.
> >>
> >> +1.  The only person saying otherwise right now is Tim, and he grants
> >> that he is only concerned with a single use case.
>
> > Hm?  I’m perfectly OK with a minimal edit leaving most things as they are
> > and just pointing out potential interop gotchas.   JSON is what JSON is.
> > For a BCP doc or “Internet JSON” profile, it’s a different story.  -T
>
> I think John thought that you wanted a strict "MUST NOT have dup
> names" requirement, but if you'd just settle for an out for streaming
> generators/parsers, pushing the non-dup name responsibility up a layer
> (or three) then maybe we could just all agree.
>
> Nico
> --
>