- From: Patrick Stickler <patrick.stickler@nokia.com>
- Date: Thu, 17 Feb 2005 10:12:01 +0200
- To: "ext Bullard, Claude L (Len)" <len.bullard@intergraph.com>
- Cc: www-tag@w3.org
On Feb 16, 2005, at 16:23, ext Bullard, Claude L (Len) wrote: > > Another way to view that is that it is inherently confusing > to use http URIs as syntax walls given that overloading the > syntax more ubiquitously used for locators confuses people. > > It's a built in danger. Always was. Always will be. > My advice is: > > a) Do the right thing and put something at the location that > settles the confusion such as pointing to the documentation > of the namespace, and use a version attribute otherwise. This is not practical, because (a) too many URIs already used as namespace names do not identify namespace documents and (b) there are those, such as myself, who have technical issues with the whole concept/methodology of namespace documents and thus would oppose any mandate to do so. From the perspective of a software agent, no presumption can ever be made about what the URI used as a namespace name might identify or whether the resource identified has any relation whatsoever to the namespace -- and thus, unless the agent has specific, reliable knowledge that the URI identifies a namespace document, such that dereferencing it will provide information relevant to terms grounded in that namespace, it should not attempt to dereference any namespace name URI. Doing so would be risky at best (and IMO improper behavior not licensed by the specifications). Rather, "doing the right thing" means not presuming anything about namespace names other than they serve to syntactically differentiate between terms. Any relationship between a namespace, terms in that namespace, and whatever might be identified by the namespace name URI should be considered coincidence, and reflecting localized, proprietary usage governed by localized, proprietary methodologies and practices and thus inherently non-portable and non-scalable to the global scope. > > b) It must be W3C policy that a namespace designer publishes > the policy for updates to the namespace. The form of that > publication is left for innovation but it must be addressable. > No this has no teeth but that is true of W3C specs in general > and that's ok. The problem here is that such a policy is suggestive that the namespace is somehow equivalent to or governs the specification/use of vocabularies, schemas, ontologies, etc. It doesn't. A namespace is just a name-space. Period. And how terms grounded in a given namespace are defined and used is an entirely separate matter. What application designers care about are the models which are significant to the proper behavior of their solutions, not trivial syntactic issues. What matters is that if e.g. some version of a model e.g. XHTML or XSLT changes, that it be clear how that model has changed, and if new terms were added, how those terms are to be used *for that version of the model* and applications should focus on the versions of the models employed -- not on the syntactic naming conventions used to identify terms employed by various versions of the same model, or even disparate but related models. namespace != vocabulary namespace != document model namespace != ontology As soon as folks come to understand and accept that, these kinds of issues will (finally) go away -- and more energy will be available for dealing with real challenges (such as how to identify which model(s) should be applied when some arbitrary term is encountered (and no, the solution is not to try to dereference the namespace name of the term...) > > There are three technical decisions that will bedevil > the web architecture until the web itself is replaced: > > 1. Grandfathering HTML. > 2. Reserving the xml: namespace. > 3. Using http URIs for namespace names. > > There are sufficient reasons for all of these except > possibly item one because it is a social compromise, > but we'll pay in friction for all three. Get used to it. True. Still, we should reduce the friction and pain by being clear about what is or is not licensed by the specs rather than introducing concepts and methodologies that further exacerbate the situation. The introduction of the concept of "namespace document" and the presumptions "marketed" in relation to namespace documents and RDDL that a namespace name URI *should* identify a namespace document just adds spin to an already dizzy issue. And it simply distracts folks from finding a proper solution which honors existing namespace usage (where the namespace name URIs do not and have never been intended to identify a namespace document) and which focuses on the semantics of information models rather than the syntax of term identifiers. It's further disappointing to see so much dicussion of namespace documents in AWWW when there is no establish usage of such a methodology and (presumably) AWWW is intended to reflect the distillation of experience and wisdom arising from established and proven practices; such that readers of AWWW may be misled into thinking that such practices are either widespread or proven (but that's an entirely different discussion...) Regards, Patrick > > len > > > From: www-tag-request@w3.org [mailto:www-tag-request@w3.org]On Behalf > Of > Patrick Stickler > > I'm afraid I don't entirely follow what you are referring to here. > Break how? I would suspect that if applications are breaking, then > they are making presumptions about the significance of the namespace > that are not strictly licensed by the specs. > > Cheers, > > Patrick > >
Received on Thursday, 17 February 2005 08:15:50 UTC