Three Dataset classes are mentioned in the specification: the dcat:Dataset from which inherits the disco:LogicalDataset, dcterms:Dataset from which inherits the disco:DataFile, and the qb:DataSet linked to disco:LogicalDataSet via disco:aggregation. Those relationships are clear.
However, it is said that "may point to multiple (0 - n) variables (Variable)" through the "inputVariable" property. I can see several issues here:
The inputVariable property is not defined in the HTML specification (only in the ttl file).
The qb:DataSet deals with aggregate data. A qb:DataSet is defined in the RDF Data Cube by a qb:DataStructureDefinition which has qb:DimensionProperty and qb:MeasureProperty. Here qb:DataSet points to variables which is neither in line with the qb:DataSet usage advocated by QB authors (even though nothing pevents formally to bind it to variables or other object types) nor with the DDI 3.2 specification which defines NCubes with dimensions. Then I am wondering why Disco has left aside the NCubes which are very similar to qb:DataSet: e.g disco:NCube owl:equivalentClass qb:DataSet;
Aggregate data should be described by Data Cube. The common use of Data Cube and Disco would be future work. An example describing the use of both would be helpful.
Three Dataset classes are mentioned in the specification: the dcat:Dataset from which inherits the disco:LogicalDataset, dcterms:Dataset from which inherits the disco:DataFile, and the qb:DataSet linked to disco:LogicalDataSet via disco:aggregation. Those relationships are clear. However, it is said that "may point to multiple (0 - n) variables (Variable)" through the "inputVariable" property. I can see several issues here: