"My goal is to implement a fully-featured query system using just database stored procedures and plain ANSI-92 SQL statements."
"The prototype uses triples gathered from the FOAFnaut crawler and made available by Jim Ley. There are approximately 400,000 triples. This represents a small triple store but what is of interest is that some predicates are very popular (for example, the 'knows' predicate occurs in around 100,000 of the triples, so the work in evaluating 'least popular' terms first clearly pays dividends. Queries of 5 query terms are around 20ms on a laptop PC."
"An immediate goal is to import a much larger triple dataset, such as that used by 3store, which contains approximately 5 million triples. I also want to revisit the database schema, and ensure that existing RDF model structures are supported robustly. Once this has been achieved, I want to start to look at more sophisticated capabilities such as inference and subclasses."
Solving RDF triple queries with a relational database - another approach