rhm said:Anyone who thinks relational databases have some fundamental limit on the types of data they contain hasn't read C.J. Date. I hope this guy realises that his 'technology' could be implemented as a library over a relational database. The real question is: does making a database engine specifically for this kind of datastructure lead to greater performance? If it does, great.
But it's only going to be faster for certain kinds of data that fit the way it works, just like SQL Databases will continue to be faster for OLTP type operations. Column-oriented databases will be faster for business intelligence queries. People implementing full-text search databases will continue to make their own database backends. There are often new ways of storing data, but they are nearly always less revolutionary than their creators claim and much less widely applicable.
Statements (triples or quads) can be represented in 3- or 4-column tables. There is also a mapping from the SPARQL graph query language to the relational model and to SQL which appears to have boosted performance for graph queries in implementations. Still, the model of expression is different. I don't believe any universal claims of superiority have been made - esp. wrt. performance, perhaps only wrt. expressiveness.
