- From: Arthur Ryman <ryman@ca.ibm.com>
- Date: Fri, 25 Jul 2014 15:46:09 -0400
- To: public-rdf-shapes@w3.org
I'd like to clarify IBM's position on these topics. 1. There is a well-motivated need for a way to describe RDF resources. This includes a description of the expected triples in a resource (aka graph) and the constraints they satisfy. The W3C semantics for RDFS and OWL were about inferring triples, not checking constraints. Clark & Parsia had previously proposed the ICV semantics for OWL, but this was not submitted to W3C. There was a gap in the standards space, so at OSLC we developed the Shape spec. We used the term "Shape" because "Schema" was already used by RDFS for inferring triples. We thought "Shape" was a good term because RDF resources can be visualized as graphs. 2. We believe that there should be a high-level vocabulary for describing the most common constraints. This includes things like occurence (cardinality), domains, ranges, etc. This vocabulary should be easy to understand by typical developers. It should also be easy to consume by tools. Since any consuming tool would be dealing with RDF resources, the simplest syntax for the constraints is some RDF syntax, e.g. Turtle. We did not want to create a new syntax since that would require the corresponding development of parsers, and that would slow down adoption. 3. We believe that the semantics of the high-level constraint vocabulary should be precisely defined in order to promote good interoperability. The simplest way to define semantics is to express the meaning terms by translating them into some pre-existing language that has well-defined semantics. The obvious choice is SPARQL since it is a W3C standard, has many implementations, and defines ASK queries. The choice of SPARQL to define semantics does not imply SPARQL must be used for implementations. In fact, most of the systems that implement OSLC specifications are not natively built on RDF or SPARQL. 4. We believe that there should be an extension mechanism to define more complex constraints. SPARQL again is a good choice for the same reasons as #3. 5. Rather than invent a new mechanism for associating SPARQL constraints with RDF resources, we believe that SPIN provides a good starting point since it has been submitted to W3C, has more than one implementation, and there is a good body of implementation experience with it. 6. ShEx is an interesting approach and we are happy to see it developed further, perhaps as a transformation engine, but its use for constraint checking is not compelling at this stage since we already have more mature alternatives. Our starting position should be SPARQL and SPIN. If these fail to satisfy the use cases, then ShEx should be re-examined. That being said, we regard OSLC Shapes and SPIN as being inputs to the standardization process and expect them to be modified in the normal course of standards development. Regards, ___________________________________________________________________________ Arthur Ryman, PhD Chief Data Officer, Rational Chief Architect, Portfolio & Strategy Management Distinguished Engineer | Master Inventor | Academy of Technology Toronto Lab | +1-905-413-3077 (office) | +1-416-939-5063 (mobile)
Received on Friday, 25 July 2014 19:46:54 UTC