> -----Original Message-----
> From: ext Katz, Stephen (GILW) [mailto:[log in to unmask]]
> Sent: 30 October, 2001 16:42
> To: [log in to unmask]
> Subject: Re: [POLL] What is at the end of the namespace?
>
>
> An observation from a non-expert in namespaces...
>
> With the description of namespaces as portrayed by Patrick
> Stickler, I wonder
> how computers and automated routines are able to effectively exploit a
> namespace... In particular, for purposes of automation and to
> meet the goals
> of the Semantic Web it certainly is tempting (though perhaps not in
> accordance with the namespace specification) for applications
> to be able to
> access something useful (e.g. a formal schema definition) at the URL
> specified at the end of a namespace.
>
> However, seeing that the experts consider this as a misuse
> of namespaces,
> what is the suggested mechanism for supporting systematic
> access to schema
> definitions and/or concept descriptions by a software applications?
>
> Steve Katz
The problem is that because folks are inferring a significance to
namespace URIs that doesn't actually exist, and making very narrow,
limited applications based on that inference (such as RDDL, based
on URLs which have a 1:1 correspondence to specific ontologies or
document models -- and even there it's not perfect) there is very
little work being done on actually addressing the full breadth of
this problem.
If I create an instance of e.g. XHTML Strict, or vCard, the structure
by which that instance must conform and the semantics attached to
the structure and constructs which form the basis of interpretation
of that instance are not defined by namespace URIs. They are defined
by (possibly many different) schemata that might use names grounded
in a common namespace (for convenience) and define structural and
semantic constraints and knowledge about constructs named thus,
but the namespace itself is nothing but *punctuation* and in fact
the schemata may utilize terms from multiple namespaces, in which
case, exactly which namespace URI denotes the schema? None of them.
There need not be any 1:1 correlation between document model and
namespace. In fact, a single doctype might employ multiple namespaces
(hello! how about XML Schema folks!) or there may be multiple
doctypes which employ the same vocabulary grounded in the same
namespace(s) (hello! how about XHTML Strict and Transitional folks!).
So in fact the namespaces used to ground particular terms in a
globally unique *name* space tell us absolutely *nothing* about
how to either validate or interpret that instance! Nothing
whatsoever. Nada. Zip. Putting anything at the end of a namespace
URI (if it happens to be a URL) might be a useful hack, but it's
not a valid foundation for a global solution of relating knowledge
about models embodied in an instance to that instance or as the
identity of the models themselves.
The issue of how to relate structural and semantic knowledge to
particular resources simply is being ignored because so many folks
are in denial about the fact that all of their cute hacks based
on namespace URLs are built on top of a house of cards.
It's very troublesome, actually...
Regards,
Patrick
--
Patrick Stickler Phone: +358 50 483 9453
Senior Research Scientist Fax: +358 7180 35409
Nokia Research Center Email: [log in to unmask]
> -----Original Message-----
> From: Patrick Stickler [mailto:[log in to unmask]]
> Sent: Tuesday, October 30, 2001 2:06 PM
> To: [log in to unmask]
> Subject: Re: [POLL] What is at the end of the namespace?
>
>
> (Sorry I arrived late to the party... ;-)
>
>
> What is at the end of a namespace? Nothing. Nada. Zip.
>
> Namespaces are punctuation. Nothing more.
>
> A namespace URI is punctuation. It's sole purpose is to
> provide distinction partitions for localized names. It
> is not required to, nor should be expected to resolve to
> anything. And any expectation for it to resolve to anything
> is in error.
>
> It is true that some standards and methodologies presume,
> expect, or even require that namespace URIs resolve to
> something specific, but this is not specified by the
> XML Namespace spec, and typically only is meaningful for
> URLs (not arbitrary URIs) and such schemes fail to support
> arbitrary, yet valid, URI schemes and miss the fact that
> there need not exist any perfect intersection between
> namespace, vocabulary, ontology, or information model.
>
> A namespace does not equal a vocabulary. A given vocabulary
> can include terms which are grounded in multiple namespaces
> and can be a subset of terms grounded in a given namespace.
> The common 1:1 correlation between namespace and vocabulary
> is a coincidence of convenience, not a characteristic of
> the namespace itself.
>
> A namespace does not equal a information model (doctype). Even
> if a given document type utilizes a vocabulary which has all
> terms grounded in a particular namespace, the namespace itself
> does not denote the content models defined for the constructs
> denoted by those terms. Furthermore, a given doctype may
> utilize terms from disparate vocabularies which each have
> terms grounded in various namespaces. And finally, multiple
> document models may be defined which all share a common set
> of vocabulary terms.
>
> A namespace does not equate to a particular schema encoding.
> Even in coincidental cases where a single document model
> uses a single vocabulary where all terms are grounded in the
> same namespace, there could be multiple schema types (DTD,
> XML Schema, RELAX, RDF Schema, DAML, etc.) all defining
> constraints and characteristics of that model. Yes, RDDL
> attempts to address such coincidental cases, but such
> treatment is based on a misunderstanding of what a namespace
> is intended to represent, only works for such coincidental
> cases with perfect intersection, and is not suitable for all cases
> of namespace usage.
>
> What is at the end of a namespace? Nothing. Nada. Zip.
>
> Namespaces are punctuation. Nothing more.
>
> I agree that we desparately need a means by which we can reference
> and relate various vocabularies, ontologies, document models,
> etc. -- but namespaces are not that mechanism, and to try to
> base a solution on namespaces will be detrimental to the web in
> the long run. It is not the way to go.
>
> I think RDDL is a step in the right direction, but the "things"
> which are being described by RDDL are not namespaces, but rather
> the vocabularies, ontologies, and document models -- and it is
> their identity which must be the basis of statements, not any
> namespace URI which in reality denotes nothing but a disjunct
> naming partition.
>
> This whole misconception of what namespaces denote was born
> out of the (mis)use of URLs for namespace URIs. It is quite
> understandable for folks to expect URLs to resolve to
> *something*, and seeing URLs as namespace URIs made folks think
> that namespaces should resolve to something. URLs should not
> be used as namespace URIs. Namespaces (and vocabularies, and
> ontologies, and document models) are all abstract entities which
> cannot be dereferenced. One may dereference a schema, but that's
> not the same thing as the abstract entity which it defines, and
> which may have multiple, equally valid definitions, all using
> the same vocabulary, all grounded in the same namespace(s).
>
> A reasonable solution to this problem will only be found when
> folks give up on the expectation that namespaces resolve to
> *anything* or equate to vocabularies, ontologies, or document
> models, and begin to think in terms of a more proper means of
> defining the identity of such abstract entities and begin
> making statements about their schema-based definitions and
> the locations of those definitions for the sake of automated
> processes, in terms of those proper (non-namespace-based)
> identities.
>
> What is at the end of a namespace? Nothing. Nada. Zip.
>
> Namespaces are punctuation. Nothing more.
>
> Best Regards,
>
> Patrick
>
> --
>
> Patrick Stickler Phone: +358 50 483 9453
> Senior Research Scientist Fax: +358 7180 35409
> Nokia Research Center Email: [log in to unmask]
>
|