Almost every RDF store actually stores more than triples; they store "quads" or "contexts" or provenance or something...
The cwm formula interface used to use (f, p, s, o), where f was the formula in which the triple occured; then it grew to (f, p, s, o, why) where why was a proof structure. Another common pattern is (s, p, o, source), where source points to a file/resource that the triple was loaded from. In re-writing an RDF store for tabulator, TimBL seems to be settling calling that 4th column why, which works like source in common cases. This is inspired at least in part by some truth maintenance stuff that Chris presented Oct 6 2005 in the Notions and Notations class.
Meanwhile, Edd observes lots of submissions about RDF Stores for XTech 2006. I hope that folks building stores look at TMS architectures. At least one group is:
- Jeen Broekstra and Arjohn Kampman. Inferencing and Truth Maintenance in RDF Schema: Exploring a naive practical approach. In Workshop on Practical and Scalable Semantic Systems (PSSS), Sanibel Island, FL, 2003.
Word has it the classic papers are:
- J. Doyle. A Truth Maintenance System. In Readings in Nonmonotonic Reasoning, pages 259–279. Morgan Kaufmann Publishers, San Francisco, CA, USA, 1987.
- Johan de Kleer. An Assumption-based TMS. Artificial Intelligence, 28 (2):127–162, 1986.
See also: WebDataInterfaceDesign.