I like JSON schema, but I haven't found any good tools to designing them. Or at least not a few years ago when I was looking.
However, we already had an XMLSpy license, so decided to just stick with designing XSDs in XMLSpy and then just translate that to JSON Schema.
If you make some small decisions, like value with attributes becomes an object, you can get a fairy decent subset of XSD to map 1:1 onto JSON Schema 2020-12.
As a nice side effect of writing the XSD to JSON Schema converter, it's trivial for us to support reading XMLs and convert that to JSON. Great for the customers who have programs that doesn't speak JSON.
I think the post is generally pretty good. There are some things that I would have stated differently.
"Unfortunately, [the terms] leaked into the documentation that everyone reads"
- We did this on purpose to align everyone's terms. It makes things so much easier when the people asking and answering questions are using the same language.
There are some interesting conceptions of OpenAPI in here as well. Specifically, OpenAPI isn't a JSON Schema document. It's its own kind of document that has JSON Schemas embedded in it.
Still, it's a decent high-level summary. If you're interested in diving a bit deeper, definitely come visit us in Slack (https://json-schema.org/slack).
I'm working with JSON schema through OpenAPI specifications at work. I think it's a bit of a double-edged sword: it's very nice to write things in, but it's a little bit too flexible when it comes to writing tools that don't involve validating JSON documents.
I'm in the process of writing a toolchain of sorts, with the OpenAPI document as an abstract syntax tree that goes through various passes (parsing, validation, aggregation, analysis, transformation, serialization...). My immediate use-case is generating C++ type/class headers from component schemas, with the intent to eventually auto-generate as much code as I can from a single source of truth specification (like binding these generated C++ data classes with serializers/deserializers, generating a command-line interface...).
JSON schema is so flexible that I have several passes to normalize/canonicalize the component schemas of an OpenAPI document into something that I can then project into the C++ language. It works, but this was significantly trickier to accomplish than I anticipated.
I used to believe that I was working with JSON schema through OpenAPI 3.0, but then I learned a hard lesson that it uses an “extended subset” of it. And what does that mean? It “means that some keywords are supported and some are not, some keywords have slightly different usage than in JSON Schema, and additional keywords are introduced.” [1]. Yes, that’s a bonkers way to say “this is not JSON schema although it looks similar enough to deceive you”. This word game and engineering choice is so bizarre that it’s almost funny.
OpenAPI 3.1 replaced the not-a-superset-or-subset of JSON Schema with the actual JSON Schema (latest version) over five years ago. No one should be using 3.0.x anymore. And 3.2 came out a few months ago, containing lots of features that have been in high demand (support for arbitrary HTTP methods, full expression of multipart and streaming messages, etc).
Many users are stuck at 3.0 or even Swagger 2.0 because the libraries they use refuse to support recent versions. Also OpenAPI is still not a strict superset because things like `discriminator` are still missing in JSON schema.
If you're building a brand new, multi-language, multi-platform system that uses advanced open-api features - you will get bitten by lack of support in 3.1 versions of tooling for features that already existed and work fine right now in 3.0 tool versions. Especially if you're using a schema-first workflow (which you should be). For example, $ref's to files across windows/linux/macos across multiple different language tools - java, .net, typescript, etc.
If you need (or just want) maximum compatibility across tools, platforms and languages - open-api 3.1 is still not viable, and isn't looking like it will be anytime soon.
This and schema support for JSON will always be an ill fitting afterthought. If you really, really need the strictness and correctness (and you most probably don't) XML
has you covered.
I'm currently rebuilding my implementation JsonSchema.Net from the ground up. I completed 95% of it in a weekend, and it already supports all the published versions Draft 6 and later, including the more complex keywords. Granted, I have a particularly deep understanding of the spec already, but it doesn't feel overly burdensome.
The metaschemas are useful but not strict enough. They don't set `additionalProperties: false`, which is great if you wanna extend the schema with your own properties, but it won't catch simple typos.
For example, the following issues pass under the metaschema.
This is intentional because unknown keywords are permitted with JSON Schema 2020-12 and prior. We are changing this with the upcoming version, which means we'll be updating the meta-schema to enforce it as well.
Good introduction, sad that it's needed. I've written hundreds of schemas and massive meta schemas and used most of the 2020-12 spec. I still struggle to navigate and use json-schema.org documentation to look up simple information.
JSON Schema could get more traction if its homepage was oriented more towards users instead of implementors.
I think JSON Schema made the same fallacy as eg. OWL did: The assumption of an open world. 99% percent of the time you want to express "This message should look like this and everything else is wrong". Instead JSON-S went the way to make everything possible at the price of rendering the default unwieldy.
The deeper principle here: "stop lying about where truth lives."
I've been exploring how this generalizes beyond side effects. Every React state library creates a JavaScript copy of state that must sync with the DOM. This is the original sin. Two truths = lies.
The solution isn't better syncing, it's refusing to duplicate. The DOM is already a perfectly good state container. All you have to do is read it.
Releasing a paper (DATAOS) and React implementation (stateless, <1KB) soon. It's the architecture behind multicardz (hyper-performing kanban on steroids, rows AND columns, 1M+ cards, sub second searches, perfect lighthouse scores, zero state sync bugs). Because there's no state to sync.
I am doing final testing, packaging, integrating strip, etc. My target was before end of month. If I work hard I just might make it.
I'm releasing a bunch of things at once:
1. multicardz.com
2. a paper on using DOM as single source of user state dataos.software
3. an OSS repo of hyper performant implementation of that for React (stateless.software)
4. an OSS repo that might best be described as Tailwind for frontend behavior. (genX.software)
However, we already had an XMLSpy license, so decided to just stick with designing XSDs in XMLSpy and then just translate that to JSON Schema.
If you make some small decisions, like value with attributes becomes an object, you can get a fairy decent subset of XSD to map 1:1 onto JSON Schema 2020-12.
As a nice side effect of writing the XSD to JSON Schema converter, it's trivial for us to support reading XMLs and convert that to JSON. Great for the customers who have programs that doesn't speak JSON.
"Unfortunately, [the terms] leaked into the documentation that everyone reads" - We did this on purpose to align everyone's terms. It makes things so much easier when the people asking and answering questions are using the same language.
"The official JSON Schema website has a validator you can try: https://www.jsonschemavalidator.net/" - Would have been better to point to the actual official JSON Schema website's tools page (https://json-schema.org/tools) that lists many online validators.
There are some interesting conceptions of OpenAPI in here as well. Specifically, OpenAPI isn't a JSON Schema document. It's its own kind of document that has JSON Schemas embedded in it.
Still, it's a decent high-level summary. If you're interested in diving a bit deeper, definitely come visit us in Slack (https://json-schema.org/slack).
I'm in the process of writing a toolchain of sorts, with the OpenAPI document as an abstract syntax tree that goes through various passes (parsing, validation, aggregation, analysis, transformation, serialization...). My immediate use-case is generating C++ type/class headers from component schemas, with the intent to eventually auto-generate as much code as I can from a single source of truth specification (like binding these generated C++ data classes with serializers/deserializers, generating a command-line interface...).
JSON schema is so flexible that I have several passes to normalize/canonicalize the component schemas of an OpenAPI document into something that I can then project into the C++ language. It works, but this was significantly trickier to accomplish than I anticipated.
[1]: https://swagger.io/docs/specification/v3_0/data-models/keywo...
Many users are stuck at 3.0 or even Swagger 2.0 because the libraries they use refuse to support recent versions. Also OpenAPI is still not a strict superset because things like `discriminator` are still missing in JSON schema.
If you're building a brand new, multi-language, multi-platform system that uses advanced open-api features - you will get bitten by lack of support in 3.1 versions of tooling for features that already existed and work fine right now in 3.0 tool versions. Especially if you're using a schema-first workflow (which you should be). For example, $ref's to files across windows/linux/macos across multiple different language tools - java, .net, typescript, etc.
If you need (or just want) maximum compatibility across tools, platforms and languages - open-api 3.1 is still not viable, and isn't looking like it will be anytime soon.
Now it feels like writing a validator is extremely complicated.
IMO, the built-in vocabularies were enough, and keeping it simple would provide more value.
JSON as a format didn't win because it supported binary number encoding or could be extended with custom data types -- but rather because it couldn't.
For example, the following issues pass under the metaschema.
JSON Schema could get more traction if its homepage was oriented more towards users instead of implementors.
*even if I would prefer more transformation/conversion features that would bring it to more more a parser rather than only a validator
I've been exploring how this generalizes beyond side effects. Every React state library creates a JavaScript copy of state that must sync with the DOM. This is the original sin. Two truths = lies.
The solution isn't better syncing, it's refusing to duplicate. The DOM is already a perfectly good state container. All you have to do is read it.
Releasing a paper (DATAOS) and React implementation (stateless, <1KB) soon. It's the architecture behind multicardz (hyper-performing kanban on steroids, rows AND columns, 1M+ cards, sub second searches, perfect lighthouse scores, zero state sync bugs). Because there's no state to sync.
I am doing final testing, packaging, integrating strip, etc. My target was before end of month. If I work hard I just might make it.
I'm releasing a bunch of things at once: 1. multicardz.com 2. a paper on using DOM as single source of user state dataos.software 3. an OSS repo of hyper performant implementation of that for React (stateless.software) 4. an OSS repo that might best be described as Tailwind for frontend behavior. (genX.software)
they are all somewhat interconnected