Uche and I have written a bit on XForms on copia. I've recently been motivated to better articulate why I think the use of XForms, Plain Old XML (POX), and GRDDL (or faithful renditions of RDF graphs if you will) is a more robust web architecture for managing mutable RDF content for the purpose of research data management than other thin-client approaches, for instance.
Some time ago, I asked:Are there examples of tools or architectures that demonstrate support for the Model View Controller (MVC) paradigm for data entry directly against RDF content? It seems to be that there is an inherent impedance mismatch with that is needed for an efficient, documented-hosted, binding-oriented architecture for data entry and the amorphous nature of RDF as well as the cost of using RDF querying as a mechanism for binding data to UI elements.
In my experience since 2006 as a software architect of web applications that use XForms to manage patient record documents as RDF graphs, I've come to appreciate that the 'CRUD problem' of RDF might have good protocol solutions being developed right now, but the question of whether there is anything more robust for forms-based data collection than declarative, auto-generated XForms that mange RDF datasets is a more difficult one, I think.
My personal opinion is that the nature of the abstract syntax of an RDF graph (as opposed to the tree underlying the XML infoset), its impact on binding RDF resources to widgets, and the ubiquitous use of warehouse relational schemas as infrastructure for RDF datasets in databases will always be an insurmountable performance impediment for alternative solutions at larger volumes that are more robust than using XForms to manage an XML collection on a filesystem as a faithful rendition of an RDF dataset. RDF/SQL databases are normalized and optimized more for read than for write - with asymptotic consequences to write operations. An architecture that directly manages very large numbers (millions) of RDF triples will be faced with this challenge. The OLTP / OLAP divide in legacy relational architecture is analogous to the use of XML and RDF in those respective roles and is an intuitive architectural style for using knowledge representation in content management systems. GRDDL and its notion of faithful renditions can be used to manage this divide as infrastructure for contemporary content management systems.For the purpose of read-only browsing, however, RDF lenses and facets are a useful alternative. However, if you need support for controlled vocabularies, heavily-dependent constraint validation, declarative and auto-generated templating, and large amounts of concurrent data entry over large amount of RDF data, the rich web architecture backplane is just very robust in my experience and in others as well.
I had to dig into the way back machine to find the XML technologies presentation John and I were supposed to give in December of 2007 (right before my life changed forever). I need to bug him to put copies of those slides on his weblog about using XForms with schematron for real-time validation as a component of data entry.