Lists Home |
Date Index |
>De : Elliotte Rusty Harold [mailto:firstname.lastname@example.org]
>Envoye : vendredi 18 janvier 2002 15:36
>A : Nicolas LEHUEN; Paul T; Bullard, Claude L (Len); Leigh Dodds;
>Objet : Re: [xml-dev] RDDL (was RE: [xml-dev] Negotiate Out The Noise)
>At 1:06 AM +0100 1/18/02, Nicolas Lehuen wrote:
>>2) ... a description of what ? A namespace ? If RDDL
>describes a namespace,
>>then let's be careful when providing arcroles for DTD and
>schemas. The case
>>where one namespace = 1 DTD and 1 schema document for each
>>that you want to support is, like Paul wrote, a dangerous
>degradation of the
>But there's no such equality. A RDDL document can contain multiple
>schemas and DTDs per namespace URI, as many as seem useful.
That's great, but that's not what I meant. I didn't write the proper thing.
I know that RDDL allows the relation 1 namespace = n schemas. However, the
problem with RDDL resides in that there are a lot of documents for which the
assumption 1 document = 1 namespaces = n schemas is false. Namespaces aren't
meant to be used as the equivalent of DOCTYPE.
How do you handle the case of, say, a RDDL document, which contains mixed
namespaces from XHTML, RDDL itself and XLink ? How could XSV fetch the RDDL
document for the RDDL document ? Please describe me the algorithm.
>>3) It is a bit related to the scalability issue, but how do you handle
>>internationalisation ? RDDL contains human-readable text,
>that's fine, but
>>not everybody can or want to read English. So will you have
>>translation of the human-readable text in the SAME RDDL
>document ? If not,
>>where should the namespace URL point to ? The English version
>? Why ? How
>>are you going to ensure consistency between various
>translations of the RDDL
>>documents, since XLinks to resources are embedded within the
>>text ? If I have to add a new resource in the English
>version, do I really
>>have to scan the Japanese version, not understanding anything
>I read, until
>>I find the place where I suspect the resource XLink should go ?
>Since RDDL is served over HTTP, this can be solved the same way it's
>solved for HTML content negotiation: have the browser and the server
>do preferred language negotiation and serve different documents
>depending on user preference.
>>The beginning of a solution could be to drop the 1 namespace
>URL = 1 RDDL
>>document containing a mix of ALL related resources embedded
>in HTML with
>>english text. Resource description is of great interested,
>but is should not
>>be treated so lightly as RDDL does.
>>The entry document found at the URL could purely XML with no
>>human-readable documents being linked as resources, with
>>for different languages residing in different resources. Yes,
>I know, this
>>would remove the great joy of seeing something nice
>immediatly appear in
>>your browser, but it's nothing an 'appropriate' stylesheet can solve.
>This completely misses the point. There has to be something at the
>namespace URL that looks sensible when humans type the namespace URL
>into a web browser location bar. That's most of the raison d'etre for
The problem with RDDL is that it started from the question "what you we find
at a namespace URL ?". The "raison d'etre" for RDDL was this answer : "a
human readable document". Fine. Let's call it NDL, then, and describe the
namespace and its tags. But I think that by allowing resources like schemata
to be mentionned in a RDDL document, RDDL tries to solve another problem,
the binding between a resource and various other resources amongst them
schematas, which is totally orthogonal to namespaces.
A namespace is NOT the same thing as a DOCTYPE. It cannot be used as a
'type' identifier for a document, but only for the elements found in the
documents. There is not a unique namespace identifier for a RDDL document.
So how are you going the find the RDDL document for a RDDL document ? How
are you going to find the RDDL document for a WAP 2.0 document (an XHTML
DTD with WAP-specific modules) ? How are you going to find the RDDL document
for XHTML + MathML module ? In all these cases, the RDDL document cannot be
found at the XHTML namespace URL, nor at the RDDL, WAP 2.0 or MathML
namespace. Huh, sorry, the RDDL for RDDL can be found at the RDDL namespace
URL, but please describe me the generic (i.e. not specifiec to RDDL
documents) algorithm that should be used to fetch the RELAX NG schema for
RDDL if all you have is a PizzaML RDDL. RDDL is centered on namespace URLs,
but it does not provides any algorithm to find resources when the source
document contains a mix of namespaces (like lots of XHTML application,
amongst them RDDL do).
Any composite document is orphan with regards to RDDL. So is it an
interesting resource description language ? I don't think so.