Lists Home |
Date Index |
> > If we agree that XML adoption is going to drive the semantic
> > web, then the
> > benefits of smoother integration with XML far outweigh the
> > advantages of
> > annotating other formats.
> What most know as RDF, is just a serialization of the RDF
> model to XML. RDF is a graph, not just a tree.
Ugh, this is the problem with XML. It crops up at some many levels in
various architectures that it's hard sometimes to know what someone else is
talking about. (Sometimes I even wonder what I'm talking about. :-).
Actually I meant that there is an underlying philosophy to use of XML for
data interchange that will drive the way that things like web services and
the semantic web develop. RDF is a very good complement to this to the
extent that it provides new and useful features. But the current
serialization syntax is not as effective as it could be in buttressing,
rather than duplication or contradicting, this XML philosophy.
> Can you provide a small example of the same type in RDFS and
> XSD and point to the duplicated stuff?
Check out the article linked by Murray Spork. There are a few good examples.
Cardinality is the one that stuck in my head.
> > And I'm not proposing merging them
> > in the same
> > place, I'm proposing using a separate annotation file (or
> > several, for that
> > matter).
> RDDL should be able satisfy you on this aspect.
Well, I can see how RDDL could be used to reference various annotations
files. In fact, schema annotation should be a "well-known" purpose IMO,
although the fact that it isn't indicates that it is (regrettably) not that
> XSD (or vanilla XML in general if you preffer) uses
> namespaces to avoid coalisions of identical node names, or to
> group a vocabulary etc. The model behind an XML node as
> <elem xmlns="http://www.foo.org/ns" />
> is an [Element Information Item] with the following properties:
> [namespace name] "http://www.foo.org/ns",
> [local name] "elem",
> [prefix] null,
> [children] empty set,
> [attributes] empty set,
> My point is vanilla XML contains no semantics; just a common
> iterpretation scheme called the Infoset.
> On the other hand, RDF(S) adds a semantic layer on top of
> namespaces: they are used to link a resource with the RDF
> Schema that defines it. The model behind RDF holds no
> namespaces, just URIs and literals. For example, the name
> rdfs:Resource expands to the resource
> No namespaces etc. Actually, nothing to do with XML...
I agree that at some point you have to have the courage of your convictions
and attached some well-defined semantics to something. The decision to leave
the target of an XML namespace undefined always struck me as somewhere
between a smart, pragmatic choice and a total cop-out. But I don't see why
the RDF approach to namespaces couldn't be embedded in the schema annotation
approach to expressing markup semantics.
> > As far as the parser is concerned, I imagine that you could
> > extend a SAX
> > parser with a module that understands the annotations
> > (basically just the
> > RDF model expressed using a more natural XML-schema-related
> > syntax) and
> > exposes the model to the user (using RDF mechanisms, as far as I'm
> > concerned).
> I am sure most would not want to extend a parser for this task.
Sorry, my point is that you *don't* have to extend the parser itself. One
party could write a new module that consumes SAX events and fulfills this
purpose, and this could then be used in conjunction with any parser support
SAX (i.e. any XML parser).
> > I don't really see why the parser itself would need to do
> > anything besides validate the document and generate SAX events.
> Depends on what you mean "validate" ;-)
Hopefully the above text clarifies this.
> If you come up with a better XML syntax to cover the RDF
> graph, I'm sure most of us would be interested in it.
Well, I made a pretty concrete suggestion, and I thought the syntax proposed
in the Murray Spork-linked article had merit as well. The key point is that
the metadata is bound to components of the schema, not contained in a
separate schema that necessarily duplicates a lot from the original.
> Could you provide an example? I was going to argue with this
> but I suspect we use the same term "document centric" with
> different meanings. I meant that in my world, this can be
> used in and for any document; it's global because of URIs.
Well, I don't think there are any examples of what can be done with RDF-type
assertions in schema annotations rather than using standard RDF syntax,
since the expressive power is the same. The point is that the web has
achieved its massive success *because* of its document-centric nature. To
claim that RDF has an advantage because it doesn't buy into this
document-centric paradigm seems wrong to me. If we are going to build a
"semantic web", it should leverage what has made the current web successful,
not ignore it.