I have been been working for 20 years in the field of artificial intelligence, so when I first heard Tim Berners-Lee’s ideas for the Semantic Web in 1998, it was an “oh boy, this is going to be fun” moment for me
Bosworth, in his MySQL conference talk made a very good point: it is difficult for people in even one company or organization to agree on a data schema: try getting the entire world to agree. Thus the problem with the Semantic Web…
In principle, ontology backed schemas should be self describing – at least to human beings. In practice, good luck.
I have what is sort of an ‘inside out’ view of making data available for consumption by software: standard ontologies are like a “public API” for a specific field of knowledge, but the “back ends” that talk to different data sources (e.g., web sites, XML/SGML data like SEC Edger, databases, etc.) need to be built by hand.
An example: if there were a standard ontology for news stories, then news providers could provide “back ends” to sit between the standard ontology and their data source. So, if you were writing software to use news, you would have a standard API for accessing many data sources.
This fits in well with what Bosworth is talking about: “back ends” could also be written for RSS/Atom feeds. I have some related material on my KnowledgeBooks.com site on the research page.
What I like about my idea is that it can be done ad-hoc by individuals or organizations defining ontologies with corresponding “back ends” to data sources as they need them. Widely used ontologies will get more “back ends”, access to more data, and eventually become defacto standards.