Discussion about dct:publisher

[sabinem]

On your README you wrote under Other points to discuss :
Also, shouldn't the dct:publisher be named, rather than being an instance of foaf:Agent? Analogous questions can be asked for dcat:themeTaxonomy and foaf:homepage.

I would like to start the discussion with this issue:

I think the dct:publisher should be implemented similar to what the DCAT-AP does in it's shacl graph: https://github.com/SEMICeu/DCAT-AP/blob/master/releases/2.0.1/dcat-ap_2.0.1_shacl_shapes.ttl:

There an agent shape is added for it and the name is a mandatory property in it: see here (as copied from the DCAT-AP Shacl graph):

:Agent_Shape
    a sh:NodeShape ;
    sh:name "Agent"@en ;
    sh:property [
        sh:minCount 1 ;
        sh:nodeKind sh:Literal ;
        sh:path foaf:name ;
        sh:severity sh:Violation
    ], [
        sh:class skos:Concept ;
        sh:maxCount 1 ;
        sh:path dct:type ;
        sh:severity sh:Violation
    ] ;
    sh:targetClass foaf:Agent .

If this is implemented the name would be mandatory and this would answer your question above.
What do you think about this approach?

[retog]

Thanks, @sabinem for starting the discussion. Adhering to the EU DCAT-AP is certainly a good thing, not sure why the current spec doesn't mandate a name-property. The question I wanted to rise in the Readme (IIRC) meant if we shouldn't enforce a "Named resource" (aka an IRI) in order to make sure publishers can be identified across datasets, i.e. specify sh:nodeKind sh:IRI. Ideally, this IRI should be dereferenceable to more data than what is reasonably included in the dataset description.

[sabinem]

Thanks, @retog, sh:nodeKind sh:IRI and thus enforcing the publisher is specified as an IRI sounds like a very good idea to me.

What do you mean with "not sure why the current spec doesn't mandate a name-property"? Which current spec? You mean that of foaf:Agent ?

I think the EU DCAT-AP explicitly states the Agent Class, because foaf:Agent by itself, can be many things: see here: http://xmlns.com/foaf/spec/#term_Agent: In this case we don't want a person for example, but an organization: see here: https://joinup.ec.europa.eu/release/how-are-publisher-and-contact-point-modelled. And the organization should come with a name and I think also with an IRI, as you suggested.

[retog]

Hi @sabinem, by 'current spec' I was referring to eCH-0200.