Re: Is promoting RDF+XML a lost cause?

On 24/11/2004, at 3:14 AM, Phil Dawes wrote:
> Unfortunately people don't see triples (or a graph) when looking at
> RDF/XML - they see a tree, with additional nasty RDF syntax.

This has been my biggest problem with RDF/XML.  Being tree-based, XML 
would not appear to be suited to serializing a graph.

I learnt about RDF via "triples", which seems to be a more graph based 
approach.  When I then saw RDF/XML it did not appear particularly 
elegant, though it did provide a complete encoding, and I could follow 
it easily enough (though I still make mistakes writing in it).

Speaking to people who are learning RDF via RDF/XML, they seem 
perpetually confused about the graph structure, and often hadn't even 
picked up on the whole "Subject-Predicate-Object" paradigm.  The 
consistency of these problems leads me to think that RDF/XML is a very 
poor way to learn RDF.

As others have pointed out, those who approach RDF from XML tend to 
want to use XPath, XQuery and so forth.  The problem with this is that 
these tools are appropriate for processing the tree-based structure of 
XML, and hence they do not deal well with the graph structure of RDF.

The problems all seem to stem from XML programmers thinking that 
RDF/XML is a new type of XML.  My perspective is that RDF is a new type 
of data structure, and RDF/XML is just an obtuse and efficient way of 
serializing it.

So I think that telling XML programmers that they should write their 
XML to be RDF/XML compliant is just wrong.  If it is appropriate for 
them to use RDF (sometimes raw XML is much better) then they should 
build their model as a graph, and only when they need to serialize 
should it be converted to RDF/XML.

Regards,
Paul Gearon

Software Engineer
Tucana Technologies
http://www.tucanatech.com

Received on Friday, 26 November 2004 01:50:43 UTC