Learning from "competing" formats

[MikeRalphson]

• fern (new to me this day)
• Amazon smithy more documentation and code seems available for Smithy 2.0 e.g. converting to OAS
• RAML - traits and overlays already under consideration, more?

[darrelmiller]

I'll throw this https://github.com/microsoft/typespec into the mix too (formerly known as cadl).

Also, I think Steven Mizell book has some great insight here https://smizell.com/language-oriented-approach/

[rafalkrupinski]

Not a big fan of handling importing of javascript files into my python program

https://typespec.io/docs/language-basics/imports/

[SensibleWood]

Imports:

• Import TypeSpec: Another TypeSpec file you've built your model on.
• Import JavaScript: Some way you've extended TypeSpec, which you'll use in your model.

In essence, nothing ends up in your Python program - you are only importing to your TypeSpec model. You'd use an emitter to generate Python from the model that you "could" use in your implementation, if you chose to work that way.

[rafalkrupinski]

If my code generator was to use such typespec, it would need to know what functions this import brings and what those functions do

[JoanPuig]

I would suggest taking a look at: https://github.com/FIXTradingCommunity/fix-orchestra-spec, A lot of the design discussions around scenarios, parameter validation and workflows are very similar.

[handrews]

JSON TypeSchema (which I have not seen before) was brought up in an old spec issue as an alternative schema candidate:

• Use TypeSchema as alternative schema OpenAPI-Specification#2839

[rafalkrupinski]

Seems TypeSchema lacks any means of describing union types or even optional properties.

[SensibleWood]

Seems TypeSchema lacks any means of describing union types or even optional properties.

Something? Something;

☝️ The question mark denotes optional. Without, mandatory

Unions: https://typespec.io/docs/language-basics/unions/

[rafalkrupinski]

@SensibleWood typespec vs typeschema

[SensibleWood]

Ah my mistake. Apologies

[rafalkrupinski]

easy mistake

[handrews]

CDDL is a data definition option, applying to either JSON or CBOR (CBOR support would be a nice bonus). It is not in JSON form, although we could either import them as separate documents or create a JSON representation if we really wanted to. CDDL 2 includes validation, annotation, and transformation on its road map:

• https://www.rfc-editor.org/rfc/rfc8610.html
• https://www.ietf.org/archive/id/draft-bormann-cbor-cddl-2-draft-04.html#name-processing-model-beyond-val

[earth2marsh]

See also: OAI/OpenAPI-Specification#3108