xml-dev - Re: Toward the self-describing web [was: Irony heaped on irony]

[Dan Connolly <connolly@w3.org>]

"Simon St.Laurent" wrote:

> 

> [apologies for the cross-post - trying to follow Dan...]

> 

> >Change? From what? The best-practice for Web resource identifiers

> >has always been that you can use them to access some representation of

> >the

> >state of the resource they identify, no? And namespaces are

> >resources just like tech reports, images, and other sorts of

> >documents and services and such, no?

> 

> No, actually, calling namespaces pointing at schemas a best practice is an

> invitation to mudslinging on XML-Dev.



Er... I don't mean to sling any mud. I presumed I was invited to join

this discussion by Tim Bray when he cc:'d me on the message

that started this thread. I attempt to monitor xml-dev at least

occasionally, but I haven't read all of it, and I'd appreciate

a pointer to the earlier discussions you mention where best practices

for XML Namespaces were agreed.





> >Disambiguation of names is a critical feature of XML Namespaces,

> >but what really makes it powerful and useful is that it makes the

> >web of XML documents self-describing: whenever you get a document,

> >you should be able to use the namespace identifiers to figure

> >out what the author of the document meant by the vocabulary

> >of tags and attributes used in the document.

> 

> It doesn't make XML documents self-describing, sorry.  The rest of the

> infrastructure is missing.  At best, it gives an application a better

> chance of sorting out what kind of markup it has and what structure it

> should have, but meaning is still out there in the wind.



I'm happy to leave "meaning" aside, but you must grant me that

a schema is descriptive of a namespace, no? That is: if

I have a Web of documents, and in each document, each

term in its markup vocabulary refers to another document

that describes that vocabulary (possibly with reference to other

documents), such a Web is self-describing, no?



> >We have documented this since Feb '98, when XML 1.0 became

> >a recommendation despite the lack of namespace support:

> >

> >       "Therefore it is essential that when a document is written to

> >       refer to a namespace, the name space definition should be a

> >       generic resource whose instances may include schemas in various

> >       languages at various levels of sophistication.  This is an

> >       essential growth point for the web. "

> >

> >       -- http://www.w3.org/TR/1998/NOTE-webarch-extlang-19980210#Evolving

> 

> This is a note, one of a series of documents that claim no official

> endorsement from the W3C,



Right... it just explains the direction we're headed in when we

charter new work and such.



> and one whose claims are contentious.



Evidently.



> >Surely the fact that Namespace URIs are identifiers is not exclusive

> >with using those identifiers to access definitions of those identifiers,

> >is it?

> 

> Not exclusive, no.  But the recent trend toward creating documents

> dependent on such practices and the schemas at those namespace URIs is

> disturbing, especially without discussion of what is widely considered a

> controversial issue, perhaps the most controversial area in XML's core today.



I believe this issue has been discussed before, without an overall

conclusion... it seemed to get as far as "you can't do that...

you need packaging to do that" vs. "packaing is fine, but

it's not critical path to using schemas and namespaces".



I think those discussions subsided because no conclusion

was necessary for the specs that were in development at the time,

and there wasn't any running code to help decide one way or

another.



But now we have a schema implementation in development, and

I'm trying to use it to show that using namespaces

and schemas works just fine. By all means, let's do

discuss it, now that we have more specifics

to work with.



> >Packaging isn't gone; it's still on the TODO list

> >(http://www.w3.org/XML/Activity).

> 

> It's been listed since September, with no public activity whatsoever.

> Apart from being a focus for alternating impatience and depression, its

> listing on 'TODO' isn't doing very much.



Umm... you mean nobody is doing experimental design nor implemention?

Nobody's doing any hacking or anything? I sure hope they are;

as Linking finishes up and we find more time to devote to packaing,

I hope there are proposals on the table so that we don't have

to start from scratch.



> >But I don't see how it's a necessary predecessor

> >to making an XML Schema available as "definitive material" about

> >http://www.w3.org/XML/1998/namespace .

> 

> It's a necessary predecessor to describing XML Schemas' interaction with

> namespaces to be a best practice, because it seems quite likely that many

> people either:

> a) won't be using schemas at all



I don't see how that represents a problem;

http://www.w3.org/XML/1998/namespace

had no schema (i.e. nothing more than a prose description as a schema)

for quite some time with no problem. So did dublin core and lots

of other namespaces.



Could you give more details on how this scenario is a problem?



> b) will be using DTDs, XDR, RELAX, or some other schema approach in

> addition to or in place of XML Schema.



Again, I don't see how this is a problem. Making an XML schema available

at http://www.w3.org/XML/1998/namespace in no way interferes with

use of RDF schemas for http://www.w3.org/1999/02/22-rdf-syntax-ns#

nor with use of RELAX for other namespaces I've seen.



Using different formats for the same namespace is also feasible,

using HTTP content negotiation... I'm working out the details

of making both HTML and an XML schema available at

http://www.w3.org/XML/1998/namespace .



I doubt many namespace designers will find it useful to use

more than formal mechanism (DTDs, XDR, RELAX), but even

that is feasible using different MIME types or perhaps

different namespace parameters on the text/xml MIME type.



And of course, anybody who is unable, unwilling, or just

uncomfortable with depending on HTTP content negotation

is welcome to ask their users to use schemaLocation

in stead. But at W3C, we have experience using content negotiation

to evolve from GIF to PNG and so on, and so I'm comfortable using

it to evolve markup descriptions of markup vocabularies

from HTML to XML.





> Managing that requires more than pointing to a schema.



I believe we've shown by demonstration that this is not true.



Could you explain how it is that what we're doing (i.e. the

schema validator that uses namespace names to find schemas)

doesn't work?



-- 

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

***************************************************************************
This is xml-dev, the mailing list for XML developers.
To unsubscribe, mailto:majordomo@xml.org&BODY=unsubscribe%20xml-dev
List archives are available at http://xml.org/archives/xml-dev/
***************************************************************************