dcat:downloadUrl as rdf:Resource (instead of literal)

[coret]

From @nfreire:

I have started to work on NDE's dataset descriptions via the Sparql endpoint and I noticed that the distributions have the dcat:downloadUrl property with a literal value. DCAT defines the range of dcat:downloadUrl as rdf:Resource, so it would be preferable to have the values as resource references rather than literals. Would it be possible to make this change?

Reference: dcat:downloadUrl

[ddeboer]

We’re probably talking about dcat:accessURL here? We’re not using dcat:downloadURL.

Currently, we keep the source’s datatype. Do we want to always convert this to IRI?

[coret]

First of all, the dataset descriptions concerned where not made by the Dataset Register, but were imported via https://archief.nl/id/dataset/foto/2-10-62ntfoto-loda-edm-distributie which had literals instead of IRI's and used dcat:downloadURL instead of dcat:accessURL both fixed now.

https://joinup.ec.europa.eu/release/how-use-accessurl-and-downloadurl recommends:

The dcat:accessURL should be used as a direct access to a file or to a page containing further instructions. It is mandatory and guarantees the existence of descriptions for the distributions. While the dcat:downloadURL is a direct link to a file. It allows software programs to use the link to get access to the file.
If only direct download access can be provided, the URL of the data should be duplicated in both accessURL and downloadURL.

I don't think we should do this duplication to dcat:downloadURL ...

Currently, we keep the source’s datatype. Do we want to always convert this to IRI?

If you look at https://schema.org/DataDownload the sdo:contentURL has range URL. So, using literals is wrong.

Of all dcat:accessURL in our triplestore 16,309 are an IRI and 498 are literals. These literals seems to be produced by only three publishing systems: opendata.picturae.com, dc4eu.nl and goudatijdmachine.nl.

So I'd propose we change our SHACL so it's emits a warning when the sdo:contentURL is not an IRI. In the same time communicate to Picturae/Vitec and DC4EU that in 3 months distributions with literals will become invalid (we then make the SHACL report an error when providing a literal). Note: the goudatijdmachine.nl datasetdescriptions will be corrected before the next crawl. We should check if there's an IRI check on dcat:accessURL.

[coret]

Query to find distributions with a literal (instead of IRI):

PREFIX dcat: <http://www.w3.org/ns/dcat#>
SELECT ?distribution ?accessURL WHERE {
    ?distribution a dcat:Distribution ;
                  dcat:accessURL ?accessURL .
    FILTER(isLiteral(?accessURL))
}