Via this posting, New, improved semantic web - now with added meaning, I read this by Dr Mark Butler. He lists the following issues with the Semantic Web:
* Produce a better XML format,
* Demonstrate RDF/XML, RDFS and OWL solving practical problems,
* Research: an XML version of OWL, a more efficient in memory and persistent version of RDF (with locking, transactions, etc.), simple API for RDF, query interface above triples, determine if context based models (quads) is better, devise a methodology for using Semantic Web tools and compare and contrast with other existing work.
Most of these questions already have answers. Commercial companies offer transactional triple stores with the ability to query above the triple layer. There is some lack of native transactional stores (there are many that work on top of SQL databases) available in the Open Source world which is probably stifling the adoption of RDF. However, it's definately possible and gives substantial, necessary improvements to the usability of RDF. The whole layered approach of RDF is similar to the abstract relational model in databases. The separation of datatypes, for example, gives positive benefits over most SQL database implementations.
Programming APIs for RDF such as Jena and others (available in tools like Redland, KAON, Sesame, swoRDFish, etc.) are quite suitable and simple to use. I would say these are less complicated than the DOM API for XML, for example.
There are already many companies using RDF, schemas, and ontologies (maybe OWL is a little too new) to solve practical problems. Much of this work is done without much publicity. Companies like Sun or the work done by the Ontoknowledge group have not only used Semantic Web technologies but also developed methodologies on how to use these tools.
I think the only things missing are a better XML format for RDF and possibly an XML version of OWL. I'm unsure if an XML version of OWL is necessary, though.