- From: Jeremy Carroll <jjc@hplb.hpl.hp.com>
- Date: Thu, 12 Feb 2004 12:31:05 +0000
- To: Sandro Hawke <sandro@w3.org>
- Cc: Dan Brickley <danbri@w3.org>, www-rdf-interest@w3.org
Sandro Hawke wrote: >>It can be found at >> >>http://www-uk.hpl.hp.com/people/jjc/tmp/trix.pdf >> >>and will be accessible at, >> > > Why use XML as the syntax for describing triples instead of RDF/XML? > > (That is, why not just say [admit?] you're proposing a reification > vocabulary?) Reification is "de re" i.e. it refers to the statements (i.e. the meaning of the triples). The named graphs here are "de dicto" i.e. they are quoted. Thus we are not proposing a reification vocabulary (However, I have been having very interesting discussions with Chris Bizer this week in which we have been exploring arguments for de re; I hope to report at the Cannes meeting). === Also the bulk of the paper is not about named graphs but about syntax, RDF/XML's problems are completely fundamental - only a complete scrubbing and starting again can really fix the RDF over XML layering issue. The key insoluble part is property names. RDF uses an open set of properties, these becomes an open set of element (and/or attribute) tags. XML formats do not work well with an open set of tags, but expect a closed set. Moreover you can't even map all URIs into qnames so RDF/XML cannot serialize all RDF graphs. RDF/XML is a lot better than it was, but there should be a plan to replace it. TriX is an attempt, I think along with Dave Beckett's work referred to in the paper, to show what a replacement might look like. (Perhaps "complement" would be better than "replace" - RDF/XML does have strengths which are only addressed in TriX by having really quite complex, and as yet unwritten, XSLT transforms). Jeremy
Received on Thursday, 12 February 2004 07:32:25 UTC