[Json] Re: A fork in the road [was: Re: Please comment on draft-ietf-jsonschema-json-schema-01]

[John Carlson <[e-mail]>]

I thought of that, and modern day parsers like JavaScript fetch() probably
move towards an integrated parse/validate structure.  But I don’t think I
can send schema to fetch().

So if you want to show me a JSON schema library that integrates parse and
validate, great, let’s see it.  I’m not talking about a layer on top of
parse and validate, I’m talking about validating while parsing.

Nothing is precluded, I’m just talking standard practice.

I am merely separating “Value” into 3 things, parsed, valid, and invalid
value.   Whereas the original definition of “Value” was valid value.  I am
also considering unparsed value (lexeme?).
Once a value goes through semantic analysis, what is it called?  Or does
semantic analysis just turn valid values invalid?  I am merely referring to
schema validation, not semantic validation, which may deal with more than
one valid value.  I realize that schema validation may have semantic
aspects.  Ultimately, semantics is externalized through RDF, OWL, etc.  I
realize there’s probably JSON versions of these.

And yes, I realize this can probably all be done in one pass.  Good luck
with that.

John

On Sun, Jun 28, 2026 at 10:50 AM Nico Williams <[e-mail]>
wrote:

> On Sat, Jun 27, 2026 at 04:15:07PM -0500, John Carlson wrote:
> > As I tried to explain, parsing to derive values is separate from
> validating
> > those values.  Validated values are valid values.  There are also invalid
> > values.  Values that are neither valid nor invalid exist after parsing,
> but
> > not after validating.
>
> If one parses into generic HashMaps and ArrayLists or whatever, then
> yes, one can have values "that are neither valid nor invalid" (as well
> as just plainly invalid) "after parsing".
>
> But the schema should not (must not! IMO) preclude validating while
> parsing to enable parsing into specific (optimized; not optimized) data
> structures where one cannot end up having such "neither valid nor
> invalid" values.
>
> Nico
> --
>