The way I see it we actually want to serialize a datastructure. We have many potential use cases not just presentation. If I download a thousand resumes I'm not as interested in viewing them as documents as I am in consuming them as datastructures, doing filtering / machine learning / ranking / ..

The way I look at it this is as much (and probably way more) a platform to write services on top of as it is a another hole to throw resume.xml down. And for that, json is undeniably better suited.

Nobody is trying to convince you to publish books in json.