This report is potentially final. Comments are welcome at RdfRDB-chairs@w3.org.
Summary:
In October 2007, the W3C held a two-day workshop to explore current and emerging technologies for RDF Access to Relational Databases. Sixty four representatives from twenty two organizations, including Semantic Web users, tool vendors, and academics, gather to present use cases and technologies.
Presentations exposed user needs and developer technologies. Participants expressed a desire to continute to develop these technologies in an Incubator Group context. Two areas of continued work were selected:
Query federation (distributing portions of a query over different databases) was also deemed important, but was tightly enough bound to mapping that federation work should be done in a mapping incubator group. Technical mail about query relational databases with SPARQL should go to and other technical mail should go to the the RDF Interest Group semantic-web@w3.org.
Semantic Web developers have been focusing on ways to express conventional relational data as RDF. The ubiquity of machine-processable relational data makes it an attractive source of Semantic Web data. RDF's grounding in universal terms makes RDF attractive to the relational database community. Many Semantic Web systems in current development provide this utility, either through projection of relational data into triples or through translating Semantic Web queries into SQL.
On 25 and 26 October, 2007, the W3C held a workshop in Cambridge, Massachusetts, to gather information about standards work in this area. The Call for Participation was @@issued@@ on 20 July, 2007. It covered the background and goals of the workshop, and invited parties interested in attending to submit position papers.
After review, the program committee accepted 18 papers and selected a subset of them for presentation, as listed in the program. 64 people attended, representing a wide range of interest. W3C is grateful to Novartis for hosting the workshop.
Use cases from Verizon, Novartis, HP, UCB Group and the EPA provided a context for evaluating the presented technologies. They discussed a variety of architectural needs.
Boeing, and Oracle explored mapping potential at a meta-level. Vipul Kashyap made a case for treating relational mappings as a specific case of general RDF mappings. Matthew Gheen mapped SPARQL tests to run over relations created with SQL DDL.
SquirrelRDF and D2RQ use rules to map SPARQL to SQL. The Virtuoso database creates a virtual triple store, over which SPARQL queries can be run.
All of these technologies require specific mappings from relational schemas to RDF graphs. Work at the University of Texas at Austin automates the process. Telefonica I+D's SPIDERS project provides a Protegé plugin to help users with the mapping.
Semantic Sitemaps provide metadata for query direction/federation.
During the wrap-up discussion, participants re-evaluated the products of the break-out sesssion in the context of future work. Around 15 participants volunteered to start and develop W3C Incubator Groups to standardize data/query mapping and query federation as well as establish a set of benchmarks for comparing different tools. Implementor's current tools met a subset of the users' needs and the expectation was that further development and standardization would be well worth the effort.
Two Incubator Groups (XGs) will be chartered:
Query federation is considered a critical functionality for system providing query mapping functionality. However, a group examining query federation was thought to have an impractical number of dependencies on mapping so query federation issues will be addressed in the mapping XG.
Existing work: D2RQ, Virtuoso, SquirrelRDF, SPASQL
Existing work: Matthew Gheen's work
While W3C Incubator Groups require minimal W3C Team resources, this area of work is of great interest to several members of the team and is considered crucial technology for the health of the Semantic Web. Draft charters are being written and discussed with members of the W3C Team.