Do we need link-catalogs for schemas? (was Re: More namespaces perversion)

[Rick Jelliffe]

Peter Murray-Rust ¼g¹D¡G

> At 08:43 08/10/98 -0700, David Brownell wrote:
> [...]
> >
> >Perhaps what we need is XML Processing Instructions to associate
> >with the namespace declarations, which identify how the bindings
> >between namespace URIs and semantics are provided ... ;-)
> >
> This *used* to be part of the namespace proposal about 10 web years ago,
> with the src= attribute. src= mapped to a schema. A schema was deliberately
> undefined. SIG opinion was against anything like this in the namespace
> proposal and it was removed.

Perhaps it is time to (attempt to) ressuscitate XML-Bind. This was a project of
mine which I put forward during the Namespaces debate. The thrust of this was that
the then namespace draft conflated two things:
    * some way to *bind* GIs into public identififiers (URL+GI or whatever) {this
is what the current namespace has limited itself to}
    * some way to *link* this identifier to all sorts of interesting information.
I tried to push the idea that having just a simple "src" attribute in the namespace
PI tied restricted it arbitrarily, and had the bad effect of hardwiring a single
vendor's schema.

So instead I pushed a catalog idea, and in particular that the namespace proposal
actually reflected a need for a new kind of link: linking from element types rather
than from instances on elements. One wants to name them because one wants to link
to or from them. I would like to raise this link-catalog idea again.

The link-catalog idea was that potentially every stakeholder (end-user, document
creator, browser-maker, DTD/schema creator) may want to use these type-links (for
documentation, schema distribution, etc).  It would be best if this were explicitly
marked up using XLL extended links. A document should be able to have a web of
type-related resources linked to/from it.

Now the great thing about using instance syntax for new schemas (e.g. XSchema) is
that
no new linking declaration system needs to be invented: we can define any kind of
link we need and markup the element in the Xschema instance. That way we get
type-links.

So now we are at the stage:

1) namespace proposal binds GI names to public identifiers
2) XSchmema proposal (and its ilk) can allow links from GIs to a catalog (e.g.
using a fixed attribute) *but* we need to actually do it;
4) XLL proposal allows extended links, with which we can implement a catalog
system, *but* we need to actually do it.

So I would propose that XML-DEV should design an element set for catalog entries,
it should cascade or be able to link with other catalogues. It might be a valuable
enhancement to XSchema, and indicate to the W3C schema people the direction that
people would like to go. (In particular, that there are many possible schemas, and
even natrual language documents are valuable for defined document types.)

The kind of thing I am thinking of is this:

<!-- example of a link-catalog -->
<XML-DEV:link-catalog>
    <entry id="lc1" GI="p" namespace="urn:www.w3.org/html4#p" >
        <description>Links for HTMLs P element type</description>
       <!-- links to schemas -->
        <a href="www.w3.org/html4.dtd#p"
            role="-//www.w3.org/NOTATION XML DTD//EN" />
        <a href="www.schema.net/Xschema/html4.xml#p"
            role="-//www.w3.org/NOTATION XSchema//EN" />
        <a href="www.ms.com/html4.txt#p"
            role="-//www.ms.com/NOTATION DCD//EN" />
        <a href="www.netscape.com/html4.htm+p"
            role="-//www.w3.org//NOTATION RDF Schema//EN" />
        <!-- links to documentation -->
        <a href="www.vhg.org/html4/p"
            role="-//XML-DEV//NOTATION Virtual Hyper Glossary//EN" />
        <a href="www.schema.net/documentation/EN/html-paragraphs.htm"
            role="-//XML-DEV//TEXT English Documentation//EN"/>
        <a href="www.schema.net/documentation/DE/html-paragraphs.htm
            role="-//XML-DEV//TEXT Deusche Erklaerung//DE"/>
        <!-- links to further catalogs -->
        <a href="www.schema.net/link-catalogs/html/p/link-catalog.xml#lc26"
            role="-//XML-DEV//SGML Link Catalog/EN"/>
    </entry>
</XML:DEV:catalog>

In this example there is a single entry. The description field indicates the
catalog entry relates to HTML paragraphs.  The first lot of links link to schema
definitions for HTML paragraphs, given in various notations (DTD, XSchema, DCD, RDF
Schema). Then comes documentation for it in various ways (VHG, English, German).
Finally comes a link to another link-catalog, which allows cascading or a web of
links.

In each entry, I have used an XLL href (a URL) and used an SGML FPI (Formal Public
Identifier) for the role atribute.  Anyone who wants to define a new role would
make up a new FPI for that role. XML-DEV would create a good starting set of FPIs
for the catalog:
    DTD
    XSchema
    English Documentation  (.. and so on for every language)
    link-catalog

After that, there should be some kind of new attibute in Xschema, available on all
elements, called perhaps  "link-catalog-href". This contains a URL pointing to an
XML-Dev link-catalog

IMHO this kind of thing would be a major addition to the power of XML: the W3C
efforts to define a new single schema definition language are, to some extent,
misguided and out-of-sequence in that it would be better to first set up an
infrastructure by which schemas and documentation can be located.  I am not
convinced that having multitudes of schema definition languages is a bad thing:
perhaps it is better to let companies compete and just provide an infrastructure
which allows them to provide good product without hijacking documents.

Cheers.

Rick Jelliffe


xml-dev: A list for W3C XML Developers. To post, mailto:xml-dev at ic.ac.uk
Archived as: http://www.lists.ic.ac.uk/hypermail/xml-dev/
To (un)subscribe, mailto:majordomo at ic.ac.uk the following message;
(un)subscribe xml-dev
To subscribe to the digests, mailto:majordomo at ic.ac.uk the following message;
subscribe xml-dev-digest
List coordinator, Henry Rzepa (mailto:rzepa at ic.ac.uk)