Lee Feigenbaum has written an excellent survey covering the current state-of-the-art application development paradigms for the semantic web, and provided a vision as to how he believes RDF application development could/should evolve.
My off-the-cuff, gut reaction is that RDF, as the data-format basis of a distributed application development framework, needs a well-defined object model to be successful. The type of flexibility Lee is advocating comes at the cost of a relatively complicated (client-side) application development paradigm. My guess is that, as least for a while, successful RDF webapps will be written against a specific store with specific schema, perhaps gleaning extra data from elsewhere, but still specialized (e.g. this supposition is made by the promising ActiveRDF framework). My gut also tells me that writing client-side CONSTRUCTs to massage data into an app-defined model may be a barrier to widespread adoption of such a model, but it’s definitely powerful, so who knows?
Otherwise, I believe in the dream and the vision, and I definitely recommend that anyone interested in RDF development should check out Queso. Very promising server-side component to this architecture.
