R: Explaining why we use RDF instead of just XML (tools)

>I am relatively new to RDF but Im not sure I understand the comparison.
>Isnt RDF all about the graph. The subject, predicate and 
>object and what
>you can do with them.  OWL is a standard for describing specific
>subject, predicates and objects for ontology representation and
>inference.  XML just happens to be one syntax for representing RDF but
>XML does not seem to be the important part of RDF (not withstanding the
>common serialization, transporting, parsing etc).  The logical
>capabilities of RDF do not seem to have anything to do with XML.  Once
>again, I am relatively new to RDF but this is what I gathered from a
>bunch of reading.  Am I missing something?
>
>Keith

After years watching this list, I'm really glad to read this statement. At
the 'beginning' of RDF development there was instead a lot of confusion
between model and its serialization.

Unfortunately the day by day reality (in my unlucky experience) is this one:

Standard organizations works (often with little budget and time) to the
definition of XML messages to be exchanged between application. So, they
concentrate more on the syntax than the underlying model.

The real problem is, infact (at least, my real problem in the project I am
now involved), to 'drive' the development of a new XML Schema in order to
have, at the end of the work, all the bricks you need to derive an RDF
model.
I mean, to have an answer to questions like, for example: 
- be sure that all the elements in the XML document, that could be a subject
in a triple, have their URI
- use URI when possible and not strange 'UID' attributes ranging over
internally defined dictionaries
- to define (well) controlled dictionary, but allowing for their future
growth
- have an XML syntax easily trasformable in a rdf+xml one, possibly just
thru XSLT 
- to allow a hierarchical representation that however underlies an RDF
relational structure
- find he correct freedom in choosing between elements and attributes
- allow optional elements and attributs, but still obtaining a 'validating'
schema
- allow vague datatype, but still obtaining a 'validating' schema
- leave some hook for the substitution of a value with a rdf:resource
attribute
etc.

However, you cannot to use violence to your XML developers imposing a too
open and relational view of the world: they have  very practical needs, with
pretty well defined clients and servers, and they do not understand the need
of being so open to other utilisations of their schema. They need to produce
a new xsd speedly and efficiently, together with parsers, real client
application, public advertising and everything needed to fix a standard
proposal.
[I remember a situation like this in the initial P3P schema definition. Am I
wrong ?]

So our (pretty clerical :-) work should be: to monitor them, and drive their
deveolpment in the rigth direction.

So, I would ask to everybody here:
? do exists a set of 'rule of thumb' to define an XML Schema in order to be
ready for its utilisation in the RDF world?
? any list of tools to ease the transformation from an XML document to an
RDF one ?

If no, what about starting to write it ? 
That means probably: to post and collect suggestion on this matter.

---------------------------------------
Andrea Chiodi (andrea.chiodi@mail.inet.it) 
CNR ITIA Vigevano, Italy