Discussed at F2F day 2 (July 18, 2017) but not resolved. See minutes:
https://www.w3.org/2017/07/18-dxwg-minutes
Main issue was that of "nesting profiles" - efficient but complex if nesting profiles owned by others; could change in unexpected ways. Also the question of whether profiles can narrow but not expand vocabulary semantics.
Makx reports that DCAT-AP is stand-alone, copying in those elements of DCAT that it will use, and adding others.
Some comments from minutes:
maybe https://en.wikipedia.org/wiki/Composition_over_inheritance is another way to couch the issues?
antoine: relationship between profiles needs to be more flexible/fluid
dsr: More like delegation of responsibilities
phila: ODRL has inheritance but answer is prohibit/allowed with default if indeterminate
a question of use cases for knowing details about profiles in DCAT. A clean model would only require the minimum information for handing over to the profile which would be responsible for handling any sub-profiles.
phila: in this space we would need some conflict resolution
... we'd have to prove it....
alejandra: in most cases profiles would just be 'leaves' - but looking across profiles is still a need
(recording a related if old technology: http://www.rddl.org/ was a design for xml namespaces that made a simple HTML page - a bit like a 'profile' - with pointers to machine formats like DTDs, XSDs etc.)
kcoyle: Do profile definers find that they are re-using existing elements or extending.