- From: Danny Ayers <danny666@virgilio.it>
- Date: Mon, 7 Oct 2002 11:24:06 +0200
- To: "Murray Spork" <m.spork@qut.edu.au>, <www-rdf-interest@w3.org>
>Paul Prescod wrote: > >[...] > >> So don't use XSLT. Renaming elements and attributes from RSS to RDF does >> not (IMO) require the power of an infoset-based, random access, >> Turing-complete language. >> >> http://lists.xml.org/archives/xml-dev/200201/msg01629.html This looks like it should make an economical substitute for much of what's done using XSLT - I eagerly await an implementation. Having said that, although this covers most things (including RSS) I don't think it can cover all bases. That XSLT is Turing-complete says it will cover all bases (optimistically assuming there are no holes in the XML and RDF models), but in a lot of cases this might turn into a very messy approach. In particular, if the source XML contains internal/external links (grove-like) then mapping these to a graph gets tricky without real graph-awareness in the transformation language. Personally I reckon a two-stage process is likely to be the best general approach. First of all transform the tree into a graph, with arbitrarily labelled arcs mapped from links in the source. Then a correspondent of XSLT (a graph-graph declarative transformation language, specified in RDF) using the (mythical) RDFPath language is used to transform the graph. I've experimented a bit in this area (I'm sure others have too), and am reasonably convinced this stuff's doable - pretty much of the problems should already been solved in xml+rdf and graph theory. Another implementation I eagerly await ;-) Cheers, Danny.
Received on Monday, 7 October 2002 05:34:16 UTC