- From: Jeff Z. Pan <jpan@csd.abdn.ac.uk>
- Date: Mon, 17 Oct 2005 17:53:08 +0100
- To: "Dave Reynolds" <der@hplb.hpl.hp.com>, <public-swbp-wg@w3.org>
- Cc: "Jeremy Carroll" <jjc@hplb.hpl.hp.com>
Hi Dave, Many thanks for the comments. I understand that, for both points, it is not easy decision at all. As for the "pointer of user defined datatypes", although the DAML+OIL solution is a non-standard approach, it is an easier (if not the easiest) approach. Regarding the "comparison of values", I appreciate that the "eq" approach could cause performance. However, I think the implementation problems might be resolvable. For example, implementations can customise the mapsTo relations and only support the needed approximation mappings. Kind regards, Jeff -- Dr. Jeff Z. Pan (http://www.csd.abdn.ac.uk/~jpan/) Department of Computing Science, The University of Aberdeen ----- Original Message ----- From: "Dave Reynolds" <der@hplb.hpl.hp.com> To: <public-swbp-wg@w3.org> Sent: Monday, July 18, 2005 10:29 AM Subject: Re: XML Schema Datatypes in RDF and OWL > > On behalf of the HP Jena team we would like to comment on working > draft 27/4/05 of this document. > > This document is very useful and we thank the Editors for laying out > the options so clearly and comprehensively. > > 1. User defined datatypes > > Jena currently implements the DAML+OIL solution. We accept the > argument in the WD that fragment identification is a syntactic > reference in the sense of XPointer and thus that the Editor's opinion > is the best route forward. > > Should this be the consensus opinion then we will consider migrating > Jena to support it in the future but will maintain backward > compatibility with existing DAML+OIL style usages where possible. > > 2. Comparison of values > > Jena currently implements option "Primitive". We regard this as the > appropriate solution and, unlike the Editors, regard the "eq" > alternative as the least satisfactory of the three. > > The theoretical problems of "eq" (it not being an equivalence relation > due to context sensitive rounding) spill over into practical > difficulties. We support the ability to retrieve semantically > equivalent literal values in searches. If "eq" were used to define > semantic equality then there would be significant performance > implications since it would no longer to be possible to index a > canonical form of the literal value. The least impractical solution > would require storage of multiple pre-rounded forms for retrieval. Our > opinion is that the space and complexity cost implications for > implementers are too high. > > We don't regard the availability of the "eq" operator in SPARQL as a > compelling argument. On the contrary it means that users who wish to > compare across XSD type branches can do so using the *explicit* eq > operator in SPARQL. > > Dave Reynolds, on behalf of the > HP Jena team > > > >
Received on Monday, 17 October 2005 16:57:15 UTC