[Json] Notes on json-schema-language-00

[Tim Bray <[e-mail]>]

I've now had time to give the draft a read.  I think it shows promise and
would be useful in meeting needs I'm currently facing at work.  I'd like to
talk it over and if we can move it forward a bit and get significant
community support for a revised draft, would be interested in proposing we
spin up a WG to see if there's IETF consensus to be had, and in
volunteering to invest cycles to help with the process.  If anyone objects
to using this mailing list for such a conversation, please say so.
Obviously since it's on github we could file issues there if that's what
people would prefer.

Ulysse, if we do have such a discussion, please don't charge off and start
revising your draft; it's helpful to keep the document stable for a while
while we figure out whether there's consensus.

First, a question: I'm a little out of touch with the json-schema work.  Is
this designed to be a compatible subset?  Or is there any other
relationship?  Depending on the answer, we might want to consider a name
for this other than "JSON Schema Language".

Some things to discuss:

   1. There are problems with JSON terminology; "JSON text", "field",
   "name"/"value", "corresponding". But these are easy to fix.
   2. I'd consider enriching the type range a bit, we don't have to stick
   strictly with JSON's impoverished diet
   1. In 4.1, I think it'd be useful to specify integer/float, bearing in
      mind the reasonableness in behavior discussed in the intro thread.
      2. I'd add a timestamp type (RFC3339 strings)
      3. I'd really want to add an enum type (constraint on JSON string
      value)
   3. I'd introduce lots of examples early in section 4, to show how each
   of the keywords might be used.
   4. In 4.4, I suggest the URI reference resolution procedure is way too
   complicated, and needn't involve the "id" element at all.
   5. Probably error objects should allow extra fields such as line/column
   numbers and human-readable explanations of what went wrong.