rivettp
commented
5 years ago That assumption that ",all new relationship types would be defined as inheriting the parent/child relationship class currently defined in L2.ttl. " is not correct. In fact the point of separating hasSource/Target from hasChild/Parent as part of issue #137 was that if new relationship types would be defined in L2 that are not parent/child relationships then they would inherit from hasSource/Target in Base. With that explanation it seems to me that a change is not needed. Especially since the original suggestion to use “parent” and “child” came from you as being more consistent with current L2-related descriptions.
One thing that could be changed is to loosen the restrictions on L2:LegalEntityRelationship that it must have a value for L2:hasParent and L2:hasChild, to refer instead to Base:hasSource and Base:hasTarget. We could do this now or in the future as and when further subtypes of relationship are added.
Please advise.
Christoph-Gleif
Emailed comment from GLEIF
I see that hasSource and hasTarget have been defined in Base.ttl, but hasParent and hasChild are still used as the property class definition in L2.ttl.
As there will be other relationship types that can or will be added in the future that are not Parent/Child (which was the point of my original recommendation), I think that relationships defined in L2.ttl should not be restricted to only the parent/child relationships associated with accounting consolidation.
The scope of ‘Level 2’ is, after all, any number of relationship types between two Level 1 LEI entities. To think of ‘Level 2’ as only the relationship types that have currently been implemented for accounting consolidation will cause problems in the future when additional relationship types are introduced.
Adding new relationship types in the future under the current L2 semantics could be done in one of two ways:
1) Another sub-ontology for new relationship types would be created (e.g., “Level2 B”, or L2B.ttl), in which the ’new’ relationship sub-type (i.e., NOT parent/child), or
2) The new relationship types would be incorporated in the existing L2.ttl
In the first case, semantic properties common to level 2 relationship semantics would need to be replicated in the new sub-ontology, but this redundancy could introduce either conflicts or unnecessary replication of relationship fundamentals defined in the existing L2.ttl
In the second case (if I understand it correctly) ,all new relationship types would be defined as inheriting the parent/child relationship class currently defined in L2.ttl. This would obviate the need for a new sub-ontology for new relationship types, but it would perpetuate the incorrect semantic notion of parent/child for relationships that did not fit that paradigm.
I think the solution is quite straightforward: simply rename the hasParent and hasChild relationship type property in L2.ttl as hasSource and hasTarget. In order to clarify the parent/child nature of the accounting consolidation relationship types, the parent/child aspect of those relationship types can be added in an annotation if desired. This simple renaming should not affect the struture of any of the object or property class definitions in L2.ttl, and it would leave open the ability to add new relationship types in the future to L2.ttl that did not conform to a parent/child interpretation.