Can JSON-LD keywords be re-defined in the @context?

[msporny]

Is this legal in JSON-LD?

"@context": {
"@type": {"@id": "http://example.org/vocab#type", "@type": "http://example.org/vocab#Class"}
}

Should we allow JSON-LD keywords to be overridden? There seem to be two options: 1) Yes, allow them to be overridden, but warn developers that bad things will happen if we allow this, or 2) Throw an exception if somebody tries to re-define a JSON-LD keyword.

[lanthaler]

It would make sense for @type, but that's it.. Overriding @id, @context, @value, etc. wouldn't make any sense.

There's a third option: Ignore those statements

[gkellogg]

We already allow them to be overridden and warn that this is probably not a good idea. A term can be any JSON string, which includes the range of keywords.

I could see some value in being able to express @type: {@container: @set}, or even @graph: {@container: @set}.

I could also see value in expressly raising an exception if an attempt is made to alias @type to an IRI or another keyword.

[lanthaler]

No, we do not allow this at the moment, we just silently ignore it. All algorithms look for direct matches and do not take context definitions into account at the moment. I could live with some ways to fine-tune how @type etc. works (think @container) but I don't really know if there are really enough use cases justifying that.

Maybe we should instead evaluate how authors could use a rdf:type property instead of @type.

[lanthaler]

RESOLVED: JSON-LD keywords MUST NOT be re-defined in the @context. If a JSON-LD processor detects that a JSON-LD keyword is being re-defined, it MUST throw an exception.

[niklasl]

The question is if this also makes it impossible to use the compaction algorithm for getting compact JSON from RDF. Consider input like:

<#me>  a :Person .

if I use a context which defines type as a regular term, like:

 "type": {"@id": "rdf:type", "@container": "@set"}

I want it to yield:

{
  "@id": "#me",
  "type": ["Person"]
}

I would at least prefer that there is no magic "@type" key in compaction if the context defines a term for rdf:type. Or perhaps by explicitly define "@type": null. Do we need an issue on compaction for that?

This need is also described in #91.

[niklasl]

The resolution to #91 addresses this need.