I'm sitting in Sea-Tac airport at 0-Dark-Thirty, waiting to go to San Diego for a week of ontology wrestling with a large auto parts company. I have my doubts about how successful we'll end up being on this, but at the end I'm expecting to have some solid tools for semantic ontology management built, including a MarkLogic based editable semantic wiki and a resource type editor.
Of late I've begun to realize that I've left the world of XML behind. It is still a close friend, of course, someone I've known intimately for years, but I think that different data formats tend to require different modes of thought about all aspects of programming. The XML mindset is chunky - you tend to think in terms of documents and narratives structures and schematic correctness, and in many places that's exactly what's needed.
The JSON mindset is nimble but all too often ad hoc - it lends itself well to short structures and requires more communication between partners about APIs, and consequently seems to constantly call for a framework that in turn creates something of a lock in to a specific approach. Useful, even highly useful, but standardization is a pain.
RDF is almost the complete opposite - it carries its own schema, but its open world assumptions shift thing away from the bag of properties approach to one where you are truly thnking at an object level, and typically at a global level as well. It requires much more up front work, and as such seems to work best for those situations where you have a lot of people dipping into the same dataset for different needs, but it is chameleon like in its ability to both consume and produce useful data (and data structures) that are largely independent of format.
Is RDF better than JSON or XML? Are oranges better than apples or bananas? It's a silly question. They are different presentations, designed for different needs and I think that in the 21st century proficiency with all of them will be required for the programmer/specialist who are increasingly dominating the prograrmming terrain.