owl:imports

[spetrac]

To understand better the importing of multiple ttl-files, like the Ontology suggests, I found that the current way of using owl:imports might not be allright and in practise introduces some errors for rdf-stores:

1. The owl ontology states that owl:imports is "The property that is used for importing other ontologies into a given ontology." and that the domain and range of this property are owl:Ontologys as well. The files that are imported in the IDS-ontology are no ontologies by themselfes, which is a semantic error.
2. The terms of an rdf-tripel are one of: NamedNode, BlankNode or Literal. The object for the owl:import tripels in the ids-ontology is denoted like a NamedNode, which must be an absolute URI. RDF-Stores are very likely to reject that files because of relative URIs.
3. Their are enough discussion on the internet about owl:imports (e.g. w3c on owl:imports or usage with rdf:about and rdf:ID), how to use them and whether they should be used as named imports or location imports, although both cases exclude the use as file-import with relative location. That makes it also unclear, if the imported file must be an ontology with same ID as the import-URI (case: named imports) or if the ontology can be found at that URI (case: location imports).

I would suggest using named imports in the main ontology and defining sub-ontologies with their own unique id. Obviously that way it is definitly harder to get the actual file for the sub-ontology, which is another problem. I understand that the owl:imports are currently very useful to merge all necessary ttl-files to one infomodel, but alternatively at least I would like to see something like file-URLs as location imports to indicate a relative file to the current one, e.g. ids: owl:imports <file:./model/infrastructure/Connector.ttl> (I know, technically relative file-URL does not exist either). At least for me it would be a quick-fix, so that the common js-rdf-stores do not through errors like "NamedNode IRI "model/infrastructure/Connector.ttl" must be absolute.". (Although this would not resolve any semantic issues with sub-ontologies.)

[clange]

Thanks for pointing this out; however, we might actually not have a problem at all. The reason for our relative-file imports such as

https://github.com/International-Data-Spaces-Association/InformationModel/blob/1e6029c0a575381017c75280b69143822fa49e4a/Ontology.ttl#L109-L110

is our internal maintenance process. We would like to keep the ontology modular by splitting it into many files. However, to generate the publishable, one-file ontology, we need to feed all these modules to the Widoco tool, which generates a one-file version of the ontology and also the HTML documentation. I just confirmed that the published ontology (source, published version) does not contain imports.

[HaydarAk]

Thanks for the clarification @clange . If there are still questions or need for discussion, feel free to reopen the issue.