Cart
[Home] [By Thread] [By Date] [Recent Entries]
I think that we now agree at 99% :). I just think we should not put an emphasis on the DTDs, and try to dissociate the concept of document type and the concept of DTDs. For me, a DTD is a schema, expressed in a particular language. DTDs should not have a special status compared to XML Schema, RELAX NG, Schematron, etc. So to me, the PUBLIC identifier is more interesting than the SYSTEM identifier, which is a URI that can be resolved into a DTD. And yes, an OASIS catalog could be used as the meta-data resource directory that we dream of, provided that we define standard URIs for a schema, a stylesheet etc. This means that we would need a way to reproduce the purpose/nature model of RDDL links, but in a single URI. Suppose that I need a RELAX NG schema, I could obtain it by resolving a URI like 'metadata://schema//RELAXNG'. URNs would be better, because they do not need to be directly resolvable. Can OASIS catalogs help resolving URNs ? Best regards, Nicolas >-----Message d'origine----- >De : Jonathan Borden [mailto:jborden@m...] >Envoyé : lundi 21 janvier 2002 18:19 >À : Nicolas LEHUEN; xml-dev@l...; Elliotte Rusty Harold >Objet : 'document types' was ...Re: Re: Flexible Schemas > > >Nicolas Lehuen wrote: > >>... You summed it up perfectly : RDDL is for namespaces only. It >> won't work for document types. So let's try to solve the problem of >document >> types together. > >To the best I can tell an XML 'document type' would refer to >that which is >defined by a 'document type definition' i.e. a DTD. A >'document type', then, >is refered to by the <!DOCTYPE declaration, i.e. a PublicID/SystemID. >Catalogs are a terrific way to associate document types with 'stuff'. > >This, of course, is mostly orthogonal to the namespace of the >root element >of a document. > >It is not completely orthogonal because XML 1.0 defines an >'element type' >simply as the name of an element (no hierarchy, inheritance etc.). The >namespace name of the root, or document, element of a document >is of course >_part_ of the element's QName. So one can correctly call the 'document >_element_ type' the QName of the document element. > >What is the relationship between the namespace name of the >document element >and its XML type? To the extent that an element is said to be >'contained' in >its namespace, a namespace can be thought of as a container >for such types. > >To be perfectly clear, however, a namespace should not be >equated with a >type. In my mind one of the most important reasons to propose >RDDL in the >first place was the fear that namespaces would be incorrectly >equated with >types. A DTD calls itself a type definition, and by extension >a schema can >be thought of as a type definition (both are properly >collections of type >definitions but nonetheless). If a namespace name were to >directly reference >a schema definition, this improper equation might become a de facto >standard, and this is exactly what we wanted to avoid. > >Regarding the 'problem of document types' either it is >completely specified >by a DTD, or else we are extending 'document type' into the >schema realm >(fair enough). Personally I think the issues related to schema >composition, >especially across schema languages is an interesting one. Some >time ago I >sketched some thoughts regarding a simple 'schema algebra' >http://www.openhealth.org/RDDL/SchemaAlgebra , ( this is not >part of the >RDDL spec), precisely as an attempt to write down some principles in a >logical fashion, and in a way that does not depend on the >intracasies of >DTDs, XML Schema, RELAXNG etc. I would say, however, that the >'problem', if >you define it as: how do we validate an XML document instance >whose type >definitions can be _derived_ from their namespaces, but >without respect to a >particular single schema. Firstly I don't know what the term >'valid' means >in such situatuions (that would be up to the specifics of the schema >language). So in general, the answer to many questions >regarding 'document >type' fall outside the scope of RDDL. On the other hand, for >the reasons I >have outlined above, at the very least it is important NOT to directly >relate a namespace to a single schema, and this is what RDDL prevents. > >> >> I hope you have understood that I'm OK with RDDL for >namespace description >> (apart a few critics). But it is only provided that RDDL is >not used for >> what is not meant, i.e. a substitute for a resource >directory for document >> types. >> > >That would be an OASIS Catalog. Designed to locate a Document Type >Definition given a Public Identifier. > >Jonathan >
|
