Lists Home |
Date Index |
----- Original Message -----
From: "Nicolas Lehuen" <email@example.com>
To: "Clark C . Evans" <firstname.lastname@example.org>
Sent: Monday, January 21, 2002 11:36 PM
Subject: Re: [xml-dev] Re: RDDL -- How would you design it?
> > Questions for the model...
> > a. Can a Document belong to more than one Class?
> This is a difficult question. Life would be much easier if the answer was
> 'no', but we should think about it and find solid reasons to support this
> > b. Can a Bundle have Resources from more than
> > one Class? If so, is Directory redundant?
> > c. Can a Class have more than one Bundles?
> Well in fact, I'd rather have a Directory per Class, with no Bundle, for
> sake of simplicity. I'd like the Directory to be referenced as an URL, so
> that different Classes from the same domain (same application or same
> semantics) could have their Directories at URL from the same domain,
> possibly with a common root path.
> The resolution of a meta-data resource will use at least three adressing
> schemes :
> 1) From a document to a document Class (depending on some information in
> 2) From a document Class to a Directory (depending on the Class URL)
> 3) From a Directory to a meta-data Resource (depending on the Resource
> URL ?)
> If we had bundles and multiple class support per Directory, the adressing
> schemes will become more complicated, and feature more stages, for example
> 1) From a document to a Directory (depending on some information in the
> 2) From a document to a document Class (depending on some information in
> 3) From a Directory to a Bundle (depending on the document Class URL)
> 4) From a Bundle to a meta-data Resource (depending on the Resource Class
> URL ?)
> I'd rather not have this extra level of complexity.
> > d. Are authorities attached to Resources or Bundles?
> > Are Authrorities even necessary?
> Like I said before, I'd rather let the authentication problem to someone
> else, not trying to solve this problem at the meta-data level only. There
> are some initiatives like XML Signature whose purpose is to provide
> for authentication and forgery protection. Let them handle this, and use
> their solutions both at the document and meta-data level.
> > Questions for an XML version of the model...
> > 1. Are Classes namespace URIs? Are all URIs a Class?
> Dunno and nope. To begin with, an URI is not a Class, a namespace, a
> document or whatever. I know by experience that it's annoying to be
> corrected on this point if you're already aware of the difference, but I
> have to make sure we agree on this point. A URI is basically a string of
> characters, specified by RFC 2396, which can be resolved into a resource
> (hence my hesitation to use Resource as the name for meta-data documents).
> Then, regarding your first question about how are Classes related to
> namespaces and namespaces related to Classes (I assume you meant that), I
> cannot reply. I just posted another mail with the subject "Finally, what
> namespaces == document types ?" that try to explain my position. Anyway,
> the *current* state of XML specs, XML Schema excepted, there are no
> relations between Classes and namespaces.
> > 2. If URIs are classes, is the document's Class the
> > namespace of the root element? Is it's collection
> > of classes (if A is yes) the collection of namespaces
> > within a XML document?
> Same thing for your first question. For your second question, I think the
> answer is definitely no, because if a document has multiple classes (which
> believe is no, but I have no reasons for the moment apart that it would
> our life bearable), and if classes were related to namespaces, I don't see
> how the set of classes of a document would be related to the set of
> namespaces it contains. I guess that if a document had multiple classes,
> classes would reside in the same namespace.
> Hum... I think this gives an idea of how a document could have multiple
> classes. A document would be said to belong to a class if it matches a
> schema associated to the class. If the document matches multiple schemas,
> then it belongs to multiple classes. This could be something to