RDF describes interrelated things in terms of subject-predicate-object "triples." Examples might be "DOCUMENT IS-A SALES_PROJECTION," or "DOCUMENT HAS-AUTHOR PAUL_SMITH." If you had lots of resources described in this way, and if the metadata vocabularies were carefully controlled, and if you had a query engine that could efficiently process sets of these assertions, you could answer all kinds of very difficult but very interesting questions. Those are three huge ifs, of course, and given the scale and chaotic complexity of the Web, it's not surprising that little progress has been made to date.
- If you had lots of resources described in this way…
- there are *huge* numbers of resources described in this way. However only a relatively small proportion of them are directly available for RDF systems. But the statements can be made available by techniques like mapping SQL database tables to RDF, or using XSLT on XML.
- if the metadata vocabularies were carefully controlled…
- the fact that folksonomic tags can be useful is just one example of where "careful control" isn't really a prerequisite. Or rather, it's not necessary to commit to using vocabularies that don't fit your requirements. Worst-case scenario, make your own up. The partial knowledge aspect of RDF means that such data can still be useful in a wider context.
- if you had a query engine that could efficiently process sets of these assertions…
- most of the RDF toolkits support one query language or another, and now SPARQL offers a standard. There are other related forms of processing (inference through RDFS or OWL, rules), but the SQL-like query is almost certain to be the biggest.
- given the scale and chaotic complexity of the Web, it's not surprising that little progress has been made to date
- massive progress has been made.
I think there's a reflection of the top-down ontology misapprehension here. If the future Web happens to look like the Semantic Web, (which I believe it is doing), then the vast majority of it will have come from localised, bottom-up efforts. It's interesting that Jon's more upbeat about WinFS. I think the assumptions he's making have led to him overlook the potential of this approach for enterprise data integration, which is something usually in scope for his writing. The distinction between metadata and data (as Jon suggest out in his introductory paragraphs) is a fairly arbitrary one. Semantic Web systems cover both kinds of music.[Danny]