- From: John McClure <jmcclure@hypergrove.com>
- Date: Mon, 10 Oct 2005 12:21:55 -0700
- To: "Daniel Oberle" <oberle@aifb.uni-karlsruhe.de>, <public-swbp-wg@w3.org>
>The primary purpose of ontologies is to enable agreement on the meaning >of specific vocabulary terms and, thus, to facilitate information integration >across individual applications. > >UML's purpose differs from that. While it can be (mis)used for this >purpose as well, it was originally intended to facilitate OOA,OOD >(please correct me if I'm wrong as I'm not an UML expert). Hi All, I am sometimes troubled by easy explanations. >From the notes about this, is it the sentiment that best-practice developers would create both a UML description of their relational database plus an OWL description of their XML external interface data? But, it seems to me that OWL now renders UML arcane in those shops creating XML/XHTML2 software from the ground-up (that is, those without legacy UML models to maintain). In fact, if it were not for RDF integration with XHTML 2.0, it's difficult to convince anyone that re-tooling for OWL-based software development is a winning investment for their scarce dollars. Pointing at applications that interface not through an API but rather through an XML entity does NOT in any way justify why RDFS/OWL ontologies are great things --- for doing so talks only to whether XML is preferred over other methods and protocols. It is really that specific issue -- OWL v XMLS -- which I think needs to be addressed front and center (plus the extent to which UML models functionally overlap with OWL models). "Agreeing on terms" can certainly be achieved using XMLS -- all non-RDF people are doing that today, which is nearly everyone!!! So why ontologies and not XML Schemas? Until that particular issue is met forthrightly, I fear that the OWL community will remain a niche while the world continues developing XML entities for interfacing applications whose element-names are indeed "agreed upon" by stakeholders. FWIW, I published a paper in 1999 which spoke to the RDFS/XMLS question, concluding that sacrificing multiple inheritance was too high a cost of XMLS; that XMLS entities are (eventually to be) transformable to RDFS entities; that there was a higher likelihood that RDFS modelling was a better fit with streamlined software development processes than XMLS (an argument as applicable to UML incidentally); and that fundamental facilities like rdf:ID did not exist in XMLS. That was before XHTML 2.0 was a reality, which I today focus on as the primary raison d'etre for RDFS/OWL. Remember, 99% of shops today create HTML applications. Tying the acceptance of ontologies by the wider community to XHTML 2.0 applications development, I think is a smart strategy commanding their attention because it addresses issues they're dealing with on a daily basis -- anything else is, well, academically interesting while not particularly compelling. Best, John McClure
Received on Monday, 10 October 2005 19:20:29 UTC