That's why RDF/XML became the biggest distraction and eventual detraction to RDF , in a nutshell. Even as of today, a majority of folks assume that RDF == RDF/XML, and that any conversation about RDF ultimately boils down to issues associated with RDF/XML :-(
Kingsley – I have to respectfully disagree. In our work to use RDF/OWL in a production environment the biggest problems were usually performance and integration into existing libraries and frameworks. Dealing with the XML serialization was rather trivial and really not that hard to manually inspect or train developers on. In the first case, we found that even though whitepaper benchmarks claimed certain levels of performance, for many of our queries and data-sets the performance wasn't the same or even useable for smallish data-sets (few 100ds GB). I'm guessing some of this was related to the nature of the test data sets versus our data sets.
In the second case, there's only so much in our use cases that we wanted from the reasoners and to do in an isolated RDF/OWL/graph-store environment. The barrier to converting OWL/RDF models to POJOs and was a real problem. With XML or JSON that represents Java type objects the integration of these objects into your systems is often trivial and there are many libraries to help (XJC, Jackson, protostuff, etc).
It's been several years since I stopped using OWL/RDF but I do periodically look for technologies that ease the integration of RDF resources in the the still dominant software development paradigm of OO objects, maps, dictionaries, and arrays.