Handle dcat:accessService

[coret]

The dcat:accessService property is a new property of the dcat:Distribution. It's meant to link to a DataService, which allows for a better description of services like SPARQL-endpoint. This new property should be allowed in the DCAT validation and stored in the right way (as we use DCAT for the data in our triplestore, a transformation is not be needed, just the right selection).

[coret]

@ddeboer could you append the selection in query.ts to handle this "extra" DCAT data?

[ddeboer]

@coret The SHACL now requires accessURL OR downloadURL OR accessService. Is this intentional? The query itself always assumed accessURL must always be available. Also keep in mind all these OPTIONALs in our SPARQL query may slow things down.

Also: accessService expects a node, not a literal (URL).

[coret]

@ddeboer the "accessURL OR downloadURL" was already in the SHACL, I now added the "OR accessService". So if the query only looks at accessURL, it should be expanded to "accessURL OR downloadURL OR accessService".

[coret]

There's an additional sub-issue. The DCAT implementation by Triply (see for example the <script> blocks within https://data.netwerkdigitaalerfgoed.nl/coret/GoudaTimeMachine) omits an dcat:mediaType for dcat:Distribution with a dcat:accessService. Currently thats required in our SHACL.

Asked about if this was an oversight or deliberate omission, the response was:

een SPARQL endpoint heeft m.i. eigenlijk geen mediatype (mime, content-type, of anders). De gebruiker van de service kan immers zelf middels een HTTP Accept header een gewenste content-type opvragen (JSON-LD, Turtle, NT, etc). Bij de schema.org hebben we het gedaan om compliant te zijn met jullie SHACL shapes, maar ook daar klopt het eigenlijk niet 100%.

I tend to agree. But this makes querying the Dataset Register also harder. Now we can just filter on dcat:mediaType (sourced from schema:encodingFormat) to find linked data. With dcat:accessService (and without dcat:mediaType) the SPARQL becomes harder. It's semantically better though.

[ddeboer]

1. How do we identify a distribution as a SPARQL endpoint, if we no longer have a format/mediaType for it?
2. While dcat:accessURL and dcat:downloadURL have a simple URL as their range, dcat:accessService has a dcat:DataService resource. How do we want to map/store that resource?