Re: [Json] Proposed rechartering for the JSON WG

[Nico Williams <[e-mail]>]

On Sat, Feb 8, 2014 at 5:10 PM, Tim Bray <[e-mail]> wrote:
> On Sat, Feb 8, 2014 at 2:50 PM, Nico Williams <[e-mail]> wrote:
>>  In absence of a JSON
>> schema language (or three) we'll have to resort to showing examples
>> and English prose for the rest.
>
> Another excellent reason not to have a JSON schema language.  I think that
> the very best protocol specs center around clear English prose and
> well-chosen illustrative examples, and schema cleverness is often a
> distraction from getting that right.

It depends on whether the schema language is easy to build tools for
or not.  ASN.1 (with all its complexity, including its encoding rules,
the IOS and the related SDL) is an example of a language which is not
easy to build tools for.

We're in a much simpler space here, not because we lack schema for
JSON but because we'd not port all the complexity of ASN.1 to a JSON
schema.  We know better.  Also, for a great many implementors the JSON
tools at their disposal (e.g., ECMAScript) make it easy to write
schema-based tools, such as code generators and schema validators.

Besides that, English prose is difficult to write in such a way that
it will be understood clearly by the many readers for whom English is
not a native/first language.  I don't ever stop to think about
non-native English readers of my prose and I'm not a native English
speaker.  I can see that English convenient for those of us who've
mastered it well enough, and it is today's lingua franca, but I'd
rather have a bit of formalism to go with the text.  Also, I
personally find some entirely-prose RFCs (like RFC4422) to be
difficult to read -- think of formalism as a crutch if you like, but
it's a very convenient crutch, helping readers, implementors, and the
authors themselves.

Nico
--