Re: on independence of elements, relating versions [XMLVersioning-41]

noah_mendelsohn@us.ibm.com wrote:

> I was trying to ask a somewhat deeper question (or maybe I'm 
> misunderstanding your answer).  Let's assume we indeed had whatever was 
> the "ideal" XML serialization of RDF, the one that most directly reflected 
> the graph of triples and was as convenient as possible for XML tools. 
> Question:  to what extent would XML -based tools then be a first class 
> means of doing RDF queries?   I still have an intuition that the answer 
> would be that there is risky level mixing going on, though I don't doubt 
> that the improved serialization would allow the XML tools to provide a 
> better approximation to an RDF query than they do today. 

I'm not sure what you mean by first class. but this argument can be 
levelled at *any* serialization of RDF - since RDF is a model. In 
that respect ther's no notable about XML that would imply second 
class means (any more than a database). That's to say there is no 
first class means. Was there some specific issue or dissonance you 
see, like graph vs tree?


> To stretch my original analogy, using XML Query on even an improved RDF 
> serializations seems a bit analagous to saying:  "if I serialize my XML 
> carefully (no comments or no CDATA sections perhaps), it will be a bit 
> easier to use Grep to reliably extract information from my files".  

That analogy doesn't follow for me. The dissonance between XML and 
grep is mainly structural. The dissonance between RDF and XQuery is 
mainly semantic. It seems to be saying more about using XQuery to 
query RDF than anything else. If you're saying XQuery might not be 
the ideal tool for querying RDF then ok, but the XML serialization 
per se is incidental to that.


>  In the same sense I 
> ask:  would XML tools applied to such an (improved) RDF serialization be 
> close to a first class means of navigating the semantic model encoded in 
> RDF, or would it be like grep over XML, an occasionally handy 
> approximation?  I'd really love to find out that the answer is "yes, first 
> class", as that would go a significant way toward relieving my concerns 
> regarding the coexistence of the two models.

XML doesn't have a model. Even if we point to the Infoset, it's not 
a model in the RDF sense, it's an abstract syntax. It's clear that 
any tool that doesn't understand the RDF model might not be optimal 
as a result. But it also might not matter.

cheers
Bill de hÓra

Received on Saturday, 20 March 2004 12:14:54 UTC