The http://www.w3.org/2008/09/rx#primarySubject predicate in RDF relates some addressable content (such as a web page, or part of a web page) to the thing which is the primary subject of that content.
For example:
<http://www.w3.org/People/Berners-Lee/> rx:primarySubject <http://www.w3.org/People/Berners-Lee/card#i>.
This RDF triple says that the primary subject of the web page at http://www.w3.org/People/Berners-Lee/ is something which is identified, itself, by the URI http://www.w3.org/People/Berners-Lee/card#i. If you click on that second link, you'll probably get some RDF data not intended to be directly read by people.
The point of rx:primarySubject is to allow RDF content to refer to things using web pages about them, instead of making up RDF-only URIs. For example, we can state the population of Bangladesh, interoperably, and without inventing any new URIs:
<http://en.wikipedia.org/wiki/Bangladesh> rx:primarySubject _:bangladesh. _:bangladesh eg:population "1613000000"^xs:decimal.
When properly defined, rx:primarySubject is an owl:FunctionalProperty, so that if two data sources both use this technique, leveraging Wikipedia, their data can be automatically merged.
One should probably use foaf:primaryTopic instead. It's the same thing, and somewhat better known.