Re: [Json] What are we trying to do?

[Tatu Saloranta <[e-mail]>]

On Wed, Jul 3, 2013 at 9:29 AM, Nico Williams <[e-mail]> wrote:

> On Wed, Jul 3, 2013 at 10:37 AM, Paul Hoffman <[e-mail]>
> wrote:
> > <no hats>
> >
> > +1 to everything Tim said, particularly "charter is self-contradictory".
> We didn't realize this when we put together the charter (a process in which
> dozens of people participated), but we do now.
>
> Re-reading the charter I'm not sure that there's a contradiction.  We
> can do something:
>
>  - remove the safe-for-eval regexp silliness
>
>  - add a section describing divergent interpretations
>
>  - leave the rest mostly as-is with only those changes for which we
> can get consensus
>
>    E.g., we can probably get consensus for "generators SHOULD NOT
> output duplicate names" and "parsers SHOULD use the last of any
> duplicate names" with appropriate text describing when and why one
> might deviate from this guidance.
>
>    E.g., we can probably get consensus for describing a subset of JSON
> strings that is guaranteed to interoperate, even though we can't ban
> unpaired surrogates.
>
> Then re-charter and publish a BCP.
>


+1 for this.

On unpaired surrogates: since their potential use has been outlined
multiple times, I am wondering if this has been substantiated by links to
systems that make use of this ability? I assume this potential exists only
on some platforms (and specifically not useful for platforms I typically
work on), but it would good to see links, similar to recent listing of
Streaming JSON processors one can find with bit of googling (happy to see
my SO answer being linked :) ).

-+ Tatu +-