RDB2RDF Deliverables

We need to produce a final report by February 2009. In addition to this report, we can also produce other artifacts such as software, papers, etc.

The XG does not create standards but can recommend that the W3C start one or more WGs to work on standards in recommended areas. This is, arguably, the most important deliverable from an XG.

Below is my attempt to summarize the conversation we had on the June 13 Telcon re. deliverables.

1. Tool and approach review:

Wolfgang Halb and Satya Sahoo have created a Wiki that surveys existing literature and tools. We agreed that this should be a deliverable from the XG, whether in its current form, updated as need be, or transformed into a paper.

Satya said we may want to publish this elsewhere such as in SIGMOD Record. Some feeling that W3C work should be published within W3C but Satya cited precedence for such non-W3C publications from other XGs.

Comment (Jenny Green): As long as we make it clear that we will publish within W3C as well as elsewhere then this should not be a problem.

2. Methodology for transforming relational data into RDF:

There was consensus that the XG recommend that the W3C start a WG to standardize a (declarative) language for transforming relational data to RDF. It was noted that this could go in two directions: from relational data to RDF and from an OWL ontology to SQL queries. (The language should be extensible to allow other features, such as rules to be incorporated into the mapping – from earlier discussions.)

Comment (Jenny Green): I think that the two directions are relational data to OWL or RDF. However I do not think that producing a declarative language means that only one of theses needs to be satisfied. It also concerns me that the mapping is always talked about in terms of mapping to SQL queries. There are specific cases in our work in which we would be required to call a web service, which would in turn call on the data in the relational database, to generate the output. I can give further explanation of our use case if the group would like more detail.

To be effective the XG should publish requirements for the mapping and, at least, the outlines of a charter for the WG

3. Standard for publishing DB/RDF update logs:

Soeren Auer described this as a facility that would indicate which parts of a (large) collection of RDF had been updated. This would help keep Search Engines up-to-date. There was some concern that this would take the XG away from it’s main focus. Soeren created the LinkedDataUpdateLogs Wiki page for further discussion.

4. Translating SPARQL to SQL

It was felt that the mapping of SPARQL queries to SQL should be a sample piece of software – open source, reference implementation -- that the recommended WG should produce to demonstrate the RDB to RDF mapping.

Comment (Jenny Green): A demonstrator is a great idea for the WG and should be included in the scope of the group to decide on the best implementation to show off their output. We have some great real world use cases that we can offer as examples for the demonstrators.

5. RDF Representation of RDB Schemas

This could be published as a W3C Note but there was no consensus to do so. We will revisit this item after the presentation by Paul Tyson on this subject on June 27th.