Question about directValueType

Holger,

Is the second sentence under sh:directValueType,  "The values of sh:directValueType must be classes (instances of rdfs:Class)." intended to be advice to people writing SHACL schemas or is it intended to impose a requirement on the SHACL engine itself (e.g.  The engine MUST raise an error if the target of sh:directValueType cannot be confirmed to be an instance of rdfs:Class or something similar)?   If it is the latter, is it the data graph, the shapes graph or both that are used to make this determination?

Harold Solbrig