RE: What to do about namespace derived URI refs... (long)

Yes, but I decide to define the serialization for instances
utilizing a particular taxonomy defined for namespace
"urn:foo:xyz" and the schema's MIME content type defines
fragment references as being hierarchical for the taxonomy
(it is a taxonomy schema) separating path elements with '::',
so I have URI refs such as "urn:foo:xyz::a::b::c". Then
later, someone else wants to use an XML Schema to define
the taxonomy for the same namespace (cause they don't like
my schema tool ;-) and then *their* URI refs become something
like "urn:foo:xyz#c" (no path, cause 'c' is a top level name
in the schema).

Now, we have one namespace, "urn:foo:xyz", but now two ways
we have to reference the same taxon "c". This becomes a
problem particularly when we must map a serialization, encoded
in some format, according to some schema formalism, into
RDF triples.

What is the common representation of name "a/b/c" within
namespace "urn:foo:xyz" (and what if we later have "a/d/c".
How do we ensure the clear distinction between the two
instances of 'c'?

Basing abstract resource identity on URL fragment references
is just wrong, wrong, wrong. Even though it has been made
to work up to now in some cases, it will bite us badly in
the behind as we progress towards any semblance of the 
semantic web.

Patrick


> -----Original Message-----
> From: ext Sean B. Palmer [mailto:sean@mysterylights.com]
> Sent: 06 June, 2001 19:20
> To: Patrick.Stickler@nokia.com; www-rdf-interest@w3.org
> Cc: Ora.Lassila@nokia.com
> Subject: Re: What to do about namespace derived URI refs... (long)
> 
> 
> [...]
> > But if two different schema encodings are used to reify the same
> > concepts, belonging to the same conceptual namespace, but which have
> > different URI fragment syntaxes, then the same RDF statements which
> > are valid according to one schema encoding are not valid according
> > to the other.
> 
> What do you mean by different fragment syntaxes? Let's take XSD once
> again as the example:-
> 
>    http://www.w3.org/2001/XMLSchema
> 
> This is the namespace that you use in all applications which do not
> concatenate names onto the end of the URI, but merely recognize the
> names as somehow belonging to the namespace as a name. For example,
> all XSD processors use this namespace.
> 
>    http://www.w3.org/2001/XMLSchema#
> 
> This is the namespace that you *could* use as an alias so as to form a
> QName that resolves to a URI that is explicitly set out in the XSD
> specification as being the intended URI reference for any of the XML
> Schema datatypes. As DanC pointed out to me, if you were using the
> following datatype in an RDF application:-
> 
>    http://www.w3.org/2001/XMLSchema#int
> 
> Any of the following would be O.K.:-
> 
>    @prefix a: <http://www.w3.org/2001/XMLSchema#> .
>    @prefix b: <http://www.w3.org/2001/XMLSchema#i> .
>    @prefix c: <http://www.w3.org/2001/XMLSchema#in> .
> 
>    a:int :name "int" .
>    b:nt :name "int" .
>    c:t :name "int" .
> 
> An XSD processor wouldn't be expected to process this, because,
> obviously, it's RDF. Also, and RDF processor wouldn't be expected to
> process an XSD file using the namespace without the hash, so there is
> no conflict.
> 
> The only conflict arises when there are no standard URI References set
> out in a specification. This is a problem with, for example, XHTML.

But what about e.g. controlled vocabularies that have no explicit
definition in any schema formalism but are valid and standard
vocabularies. Case in point ISO 3166-1 country codes. I need to define
a property that is restricted to a union of ISO 3166-1 two-letter
country codes or one of the following tokens: global, north_america,
south_america, europe, aftrica, asia_pacific, other. Where is my
authority for the ISO 3166-1 names in a schema, and if it is in
a schema, it is an enumeration, and references to enumeration values
(if that is even possible with XML Schema, I haven't clarified that
yet) would be specific to XML Schema, yet surely, the ISO 3166-1
defined country "us, United States" is not specific to XML Schema,
nor to one specific XML Schema (some other XML Schema may define
it as an empty element!).

Thus, tying the identity of abstract concepts which will serve as
the intersection of knowledge on the semantic web (which in essence
*is* the semantic web) to MIME content type specific fragment
reference syntax is just plain wrong. Even though it would be 
possible to define RDF statments to say this URI ref is the same
thing as that URI ref, etc. the combinatoric explosion of equivalence
classes would be devastating to any real-world system (and it is 
entirely unnecessary). 

Let's all step back a bit from this HTTP/URL/HTML-centric perspective and 
try to see the forest rather than those few trees.

Eh?

Patrick

PS: I'm not trying to start a flame-fest here, unless of course it
is a productive one ;-)

Received on Thursday, 7 June 2001 02:54:03 UTC