Re: [Json] [apps-discuss] JSON mailing list and BoF

[Tim Bray <[e-mail]>]

Options for \uxxxx
- forbid it
- argue that comparisons MUST be Unicode-codepoint by Unicode-codepoint,
and specify that \uxxxx specifies a single codepoint

Latter is probably better. -T


On Tue, Feb 19, 2013 at 2:08 PM, Matt Miller (mamille2)
<[e-mail]>wrote:

> [Thank you Paul; removing apps-discuss@]
>
> On Feb 19, 2013, at 3:00 PM, Francis Galiegue <[e-mail]>
>  wrote:
>
> > On Tue, Feb 19, 2013 at 10:53 PM, Tim Bray <[e-mail]> wrote:
> >> I would argue that normalization should be out of scope.  I.e. the two
> forms
> >> of “foó” are different strings, that’s all.
> >> “Doctor! It hurts when I do this!”
> >> “So, don’t do that!”
> >>
> >
> > (as a not very well educated individual on this matter...)
> >
> > If _that_ is what normalization is about, then I strongly condemn it
> > being even a subject at all.
> >
> > The current RFC does a pretty good job at defining what a JSON String
> > can be, fabricating some sort of injective function so that one JSON
> > value is equal to another is sick.
> >
> > OK, I say that, but on the other hand, JSON Schema mandates that JSON
> > values "1.0" and "1" be considered equal for numeric validation. This
> > could also be viewed as a form of normalization...
> >
>
> And furthermore, "foó" versus "\u0066\u006f\u00f3" (or
> "\u0066\u006f\u006f\u00b4"?).  I think most of us clearly think the former
> is better than the latter, but it does need to be called out.
>
> I also do think some level of discussion on unicode normalization, even if
> the result is "this is out-of-scope," needs to take place.
>
>
> - m&m
>
> Matt Miller < [e-mail] >
> Cisco Systems, Inc.
>
>