Domain of dcat:hadRole

[andrea-perego]

@dr-shorthair , I see that the domain of dcat:hadRole is prov:Influence, although in the examples we have it is always related to the "influence of an agent" (which should correspond to prov:AgentInfluence).

I guess this is meant to complement what not covered by prov:hadRole (#79 (comment)). But I wonder whether in the note to that property it may be worth making it explicit, and saying that, in the framework of DCAT, it is supposed to be used with prov:Attribution (a subclass of prov:Influence via prov:AgentInfluence), as also explained in Section "Attribution Roles".

Also, IMO it would be useful to add a link to Section "Attribution Roles", which provides examples on how to use it.

If this makes sense, I can take care of creating the relevant PR.

[dr-shorthair]

I'm also anticipating the need to assign roles to dataset-dataset relations, and don't want to have to create yetanother predicate for that case.
I started the work here: #613

Dataset-dataset relations are not attributions or generations, but something else,
I think Influence is the most general case from the PROV-O ontology, though maybe it is still too restricted as it is only concerned with causal relationships?

[dr-shorthair]

In https://github.com/w3c/dxwg/pull/651/files#diff-3536ecfed0957ecf71ed071887330f5a I propose that the domain of dcat:hadRole be the union of prov:Attribution and dcat:Relationship.

[davebrowning]

Above proposal merged in #651

[davebrowning]

@andrea-perego - does the current document now address your original concerns, or is there more to be done here?

[andrea-perego]

Yes, thanks.