- From: Danny Ayers <danny666@virgilio.it>
- Date: Sat, 2 Nov 2002 01:22:41 +0100
- To: "Bob MacGregor" <macgregor@ISI.EDU>, "Dan Brickley" <danbri@w3.org>, <www-rdf-interest@w3.org>
> >Our main feature this week on XML.com is from well-known XML > >experts Bob DuCharme and John Cowan. In "Make Your XML > >RDF-Friendly" they describe how to structure XML documents so > >that they can be used by RDF processors. > >Very nice article. However, I was hoping to read some do's and don'ts >for structuring XML so that an XML-to-RDF converter could do a good >job of converting the XML into RDF. What I think the authors were >describing >is how to structure your XML so that a standard RDF parser could >read it successfully. Some of the advice was innocuous: how to >include URI references in a compatible way. But then they started >advocating insertion of RDF-syntax (as opposed to RDF-style). The >ugly RDF:resource tag was recommended. Hmm - well it appears more like how to make your data *be* RDF, rather than just RDF-friendly. I wonder how far towards their respective cheeks the good gentlemens' tongues were... It's certainly a good article, and those 8 points make a pretty good best-practice checklist for anyone going anywhere near RDF. And then "striping". RDF >striping is an unfortunate hang-over from the bad old days that ought >to be declared deprecated. I'm sorry, have I missed something here? Apart from being perhaps a little too XML-like in appearance (thus risking misinterpretation by humans), my impression of the striping was that it was actually a thoroughly handy technique. Expecting that XML users will refactor >their XML to conform to obscure RDF conventions strikes me as >dangerously naive (dangerous to the future health of RDF). I'm not sure about this - if you rephrase that to 'represent their data so that it conforms to RDF conventions', it starts to sound very important (in a positive sense) to the health of RDF. That a lot of existing XML users might resist such a suggestion would I suppose be a good case in favour of articles like the one in question. >IMHO the RDF world should be learning how to >accomodate the XML world, not the other way around. The core good ideas in >RDF are URI's and triples (while current RDF syntax is a huge liability). >If something in XML can't be converted into triple form, >then we have a problem that needs to be handled one way or another. I think my list of core good ideas in RDF would be somewhat longer, in this context I would definitely include the graph model, further up the list than triples. All the same I don't deny that the ability to represent the graph in triples is a piece of magic. >If someone were to write an triple-compatibility verifier for XML that >indicates how successfully a given piece of XML might be converted into >RDF triples, that would be a nice contribution towards a future tool suite >intended to provide a robust means for importing XML into RDF >triple stores >(perhaps >by-passing RDF syntax completely). The problem here isn't really the transformation validity, as one could for instance use something like the Infoset RDF Schema [1] to convert any XML to RDF in a reasonably robust fashion. The transformation from (messy) tree to (bitty) graph is relatively straightforward, but following a topological approach can make for pretty convoluted semantic mappings. There may also be potential problems with transformations that don't 'understand' the model, in the same way that a dedicated XML parser is usually to be preferred over a regexp hack - things can spill over the sides. The critical piece IMHO is how well the model behind the data (*before* it gets converted to XML) maps to an RDF graph. Ok, so a lot of people might be drawn into the fold if there were nice simple XML to RDF tools, (which <!-- stretch --> isn't much of a different situation than a year or two back with RDBMS and XML). But perhaps also there may be opportunity to go for the jugular, with an RDF-modelled database which could supplant RDBMS/XMLDBs (TAP?). The producers of the 'Sentences' [2] database make some bold claims with regard to efficiency in modelling and performance for traditional DB jobs, and their model is noticeably similar to RDF. [drifts off into late night ramble-space...] A suggestion for the authors - perhaps a follow up could describe how RDF can help with down-the-wire serializations, get some more SOAP folks on board ;-) Cheers, Danny. [1] http://www.w3.org/TR/xml-infoset-rdfs [2] http://www.lazysoft.com/
Received on Friday, 1 November 2002 19:33:24 UTC