Why is node embedding not allowed for @type values (in the rdf:type - sense)

Hi All,

I stumbled upon a restriction on the allowed values for the @type keyword in a node object when specifying the rdf:type for a resource in JSON-LD.

I would expect [4] to be equivalent to [5], but where [4] is allowed, [5] is syntactically incorrect. (NB: the Python rdflib library does not complain; but jsonld.js and pyLD do)

It looks like embedding [1] (node objects as property values) is not allowed in this case, even though this is not explicitly mentioned in the section on embedding.. This is also not made clear in the section on specifying the type [2] (it just gives examples of using an IRI string as value, or an array of such strings, or a term defined in the active context).

Both sections are non-normative (which may explain things), and the restriction is only made clear in the JSON-LD grammar [3] for node objects:

If the node object<https://www.w3.org/TR/json-ld11/#dfn-node-object> contains the @type key, its value MUST be either an IRI reference<https://tools.ietf.org/html/rfc3987#section-1.3>, a compact IRI<https://www.w3.org/TR/json-ld11/#dfn-compact-iri> (including blank node identifiers<https://www.w3.org/TR/rdf11-concepts/#dfn-blank-node-identifier>), a term<https://www.w3.org/TR/json-ld11/#dfn-term> defined in the active context<https://www.w3.org/TR/json-ld11/#dfn-active-context> expanding into an IRI<https://tools.ietf.org/html/rfc3987#section-2>, or an array<https://infra.spec.whatwg.org/#list> of any of these. See § 3.5 Specifying the Type<https://www.w3.org/TR/json-ld11/#specifying-the-type> for further discussion on @type values.

My main question is: why this restriction? It is perfectly normal in RDF to have nesting for the rdf:type property (e.g. relate to a class that is itself a subclass of another class).

One reason I can think of is that the value for @type should never be a Literal; that makes sense, so any string value for the @type key will have to be interpreted as an IRI. However, disallowing for embedded node objects, seems overly restrictive to me, and I'm inclined to see this as a bug in the spec.

I can work around this by mapping a custom "type" entry in a context to rdf:type, but I'd much rather use the built-in @type for consistency.

Could someone explain the reasoning behind this?

Thanks,

Rinke


[1] https://www.w3.org/TR/json-ld11/#embedding
[2] https://www.w3.org/TR/json-ld11/#specifying-the-type
[3] https://www.w3.org/TR/json-ld11/#node-objects
[4] https://tinyurl.com/2jx8e6r4
[5] https://tinyurl.com/mr3t75zn
--
Rinke Hoekstra
Lead Architect – Knowledge
Industry Director of Elsevier’s Discovery Lab
ELSEVIER - Amsterdam
r.hoekstra@elsevier.com<mailto:r.hoekstra@elsevier.com>

Emails can arrive at all hours, but at Elsevier we respect your personal time. Feel free to respond to this email during your normal working hours.

________________________________

Elsevier B.V. Registered Office: Radarweg 29, 1043 NX Amsterdam, The Netherlands, Registration No. 33158992, Registered in The Netherlands.