Re: How namespace names might be used

Jelks Cabaniss wrote:
> 
> Michael Mealling wrote:
> 
> > I have never, ever suggested that always retrieving the resource should
> > be the expected behavior. ...
> 
> But Mr. Connolly has (as has, if I mistake not, Mr. Berners-Lee):
> 
> JR> Specifically, the new approach of putting XML Schema files at
> JR> locations specified by W3C namespace names should be immediately stopped
> JR> until there is a consensus about this.
> 
> DC> Again, why not? It works, and it's useful.
> 
> This "why not" was asked _after_ Rick Jelliffe's to-the-point explanation[1],

OK, I'll re-visit that message to see if it provides any
convincing arguments...

| But isn't the reason people don't want to think of a namespace as a
| resource because people they see it as a slippery slope to
| proprietorizing XML: when Microsoft said that BizTalk namespace URIs
| locate a schema in a particular format, that means that in order to
use
| a Biztalk document one has to buy into the schema language (i.e., the
| products) that it requires. 

No, it means that you *either* have to use the schema language
that the namespace designer chose (which may require using
certain products) *or* you use schemaLocation or any number
of other mechanisms to find a schema of your own choosing.

How is being *prevented* from using the namespace URI to
find the namespace designer's default schema better than
being allowed to?


| About a year ago, I wrote an posting to XML-DEV that "Namespaces is
| dead" trying to alert people to the problem of equating namespace with
| schema 1:1. As long as W3C prevaricates with variations on "the
| namespace URI should/can locate something which can/should
| define/describe the language", the chances of getting a workable, open
| system decreases. 

(a) I see noone claiming that namespaces and schemas are 1:1,
and (b) the rest of the paragraph is just argument by assertion,
which is a game I can play too: "the namespace URI should locate
something whish should define the language" is a powerful
way of creating a self-describing system that will revolutionize
the Web.

| We need definite ways to discover and retrieve multiple resources
keyed
| from a namepace URI: we need to be able to find if particular
resources
| relevant to that namespace in some domain are there or not there.

If I understand correctly, this suggests a protocol for asking

	Is there a Java class that implements support
	for the namespace named XYZ?

that will give definitive yes/no answers. That's not scalable.
It's like requiring link consistency in the Web. It's a non
starter: suppose we had such a protocol, and it somehow
computed and returned a "no" answer. Then, two minutes later,
I release such a Java class. How does the system learn
that it should return "yes" now?

It's perfectly reasonable to design protocols for asking

	Do you know of any Java classes that implement
	support for the namespace named XYZ?

which is exactly analagous to asking

	Do you know of any PICS labels using the java-implements
	rating system for the resource named XYZ?

and pretty much analagous to the netscape "what's related"
protocol and the other annotation protocols that TimBL mentioned
in his message of Mon, 12 Jun 2000 13:22:33 -0400
http://lists.w3.org/Archives/Public/xml-uri/2000Jun/0529.html

| Having
| a simple relationship between namespace URI and some rendition of some
| definitional entity at the other end is little use: the current
| lucky-dip system where you just don't know what kind of thing may or
may
| not be at the other end is no basis for a workable system.

More argument by assertion. Counterargument: the current robust
system consisting of URIs, HTTP, and the XML Schema language
is the basis for a powerful self-describing system.


| The constituency who is opposed to a simple retrieval from a namespace
| URI not only include those who don't feel keen on self-describing
| documents, but more importantly (and, I think, more numerously)
includes
| those whose requirements for "self-description" are not satisfied by
the
| current vague hand-waving.  

This is exactly why I believe it is so important to show, by operational
demonstration, how namespaces and XML Schemas can work together.

| This is a point that has been made many times (e.g. by Tim Bray) on
this
| list so far: so I hope when Tim BL is talking about people for whom
"the
| concept of self-describing documents just do not exist" he does not
| blanket everyone who is against the simplistic method he has been
| putting forward.

| To the contrary, I would say that the biggest thing holding back the
| development of a "semantic web" at the moment is the failure of
| architecture at the W3C: that failure being that people need
conventions
| to allow multiple resources in different domains to be retrieved,
keyed
| by the namespace.  

But... is it necessary to prevent deployment of a simple method until
richer methods are developed? History suggests not:
It wasn't necessary to prevent deployment
of HTML until XML was developed. It wasn't necessary to prevent
deployment of FTP until HTTP was developed.


In sum, Rick Jelliffe's message makes a number of interesting points,
but none
that establishes that dereferencing namespace URIs to find schemas
(among
other methods of finding schemas) won't work.

> and _after_ other posts, especially by David Carlisle, describing in
> excruciating detail the numerous problems of "why not".

I believe I have rebutted David Carlisle's arguments.

> But ... the more fundamental question, framed even earlier by Walter Perry[2]
> about the very assumption of this whole namespaces/semantics business, has never
> been answered.

There's a lot in that message. Could you state this "more fundamental"
question
briefly, please?


[...]

> /Jelks
> 
> [1] http://lists.w3.org/Archives/Public/xml-uri/2000Jun/0335.html
> [2] http://lists.w3.org/Archives/Public/xml-uri/2000Jun/0189.html

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Tuesday, 13 June 2000 09:52:24 UTC