There is some confusion about what the term "real-world object" ("rwo") refers to. It is not clearly defined in semantic web standards, but SKOS says it refers to intangible as well as tangible entities, including concepts. This is the basis of the RDA Glossary definition: an "instance of an entity" is a loose reference to a dataset, while a "term from a vocabulary encoding scheme" is a reference to the entity skos:Concept. Note that RDA includes both datasets and value vocabularies in its definition of "vocabulary encoding scheme" (VES), which does not follow the distinction made by the W3C Library Linked Data Incubator Group.
This makes no difference to the meaning of real-world object, but may lead to or confirm a misinterpretation.
"Real-world" and "object" do not equate to "tangible" or a distinction between mental constructs and physical constructs. Instead, "real-world" indicates a separation of a "thing" that is represented by an IRI in RDF and statements about the thing in RDF triples. A rwo is what is referenced by an IRI in a triple. The predicate or property of a triple is an intangible conceptual object (an element from an element set or ontology, the third kind of IRI identified by the W3C group***).
For your example, the correct object is the "names" IRI: it is an instance of Nomen.
The full triple statement using the element is:
<http://id.loc.gov/rwo/agents/n79068416> <rdaao:P50094> <http://id.loc.gov/authorities/names/n79068416> .
That is:
This person has identifier for person this name authority.
If we de-reference the id.loc.gov IRIs we find the triple:
<http://id.loc.gov/rwo/agents/n79068416> <madsrdf:isIdentifiedByAuthority> <http://id.loc.gov/authorities/names/n79068416>
This person is identified by authority by this personal name.
The usage of "real-world object" in MADS RDF suggests that a name or other appellation is not included. This is only true if the name or appellation is given as a string value of the object of a triple statement:
This person is identified by authority by "Tolstoy, Leo, graf, 1828-1910".
Otherwise, an "authority" is a rwo in its own right. We should always remember that an IRI does not carry intrinsic semantics; semantics are accommodated in statements about the IRI, when the IRI is the subject of a triple. We see many cases where an IRI tells you about itself: in this case, whether it is an "authority/name" or "rwo/agents" - but this is mostly due to the serialization of legacy datasets. RDA element set and value vocabulary IRIs are also based on a human-readable pattern that, for example, indicates the distinction between an element set and a value vocabulary (<http://rdaregistry.info/Elements/>, <http://rdaregistry.info/termList/>) or the type of entity to which an element set belongs (<http://rdaregistry.info/Elements/i/>). However, it is unsafe to assume any semantics implied by an RDA Registry IRI in an open linked data application. RDA semantics are given in human-readable descriptions and scope notes, and in machine-readable domain, range, and other element RDF properties; all of theses statements are available when the IRI is de-referenced.
I think it would be better to use rdaao:P50411 "has authorized access point for person" or the broader rdaao:P50377 "has access point for person" as the element in the example. RDA makes a basic distinction between names/titles, access points, and identifiers; if the distinction is not required, unsafe, or not available, use rdaao:P50382 "has appellation of person". Ultimately, an application should check the semantics (especially entity definitions) of external Nomen datasets (entity/authority files) before selecting a suitable RDA element.
Note that SKOS addresses the issue of describing labels (including appellations and concept terms) with SKOSXL (SKOS Extension for Labels).
*** The three "kinds" of IRI are dataset, value vocabulary, and element set or ontology. These are incorporated in RDA as the VES for category of linked data work. The RDA Registry publishes the RDA element sets and value vocabularies, but not any RDA datasets which constitute examples of instances of entities described using RDA.
------------------------------
Gordon Dunsire
------------------------------