Re: Language = Namespace. was: How namespace names might be used

At 11:22 AM 6/21/00 +0100, John Aldridge wrote:
<aside>
>My dictionary doesn't imply that, and it wasn't the impression I intended 
>to give.  Apologies if I accidentally made a more loaded statement than I 
>intended.
></aside>

No offense taken.  Thanks for clarifying your intent.

>OK.  Does the following then represent your views accurately?
>
>There are documents which are written in several 
>(language=namespace)s.  There is no concept of the language of the 
>document as a whole (except perhaps in the weak sense of the namespace of 
>the document element),

Hmmm... maybe -- in practice I would anticipate there may be a "primary" 
language, roughly in the sense you describe, but even that is not a 
complete requirement.

I anticipate that different processors would be able to extract different 
document sub-contents based on the languages that they actually 
process.  Other language statements (if not somehow flagged 
mandatory-to-interpret) would be ignored.  The "primary" language is one 
that might be common to most processors of a document.

>  and there is certainly no concept of several distinct languages drawn 
> from elements from a single namespace.

At one level, I agree, at another I'm not so sure.  I think that a single 
namespace yields a single language.  But multiple namespaces used in 
combination might yield more languages than the number of namespaces 
used.  I could imagine three namespaces/languages:  RDF, a language of some 
logic layered on RDF, and a language to describe some entities also layered 
on RDF.  Taken together with some additional inference rules, these might 
constitute a higher-level language for reasoning about said entities.

>You are content to be able to make metadata statements only separately 
>about the individual (namespace=language)s from which the document is 
>composed; and never about a particular combination of those languages.

No... see above.

>Specifically, the concept of a DOCTYPE should be allowed to wither 
>(because there is nothing interesting to be said about a document which 
>isn't already said in the union of the metadata associated with the 
>various namespaces from which it is composed).

>As a consequence, it doesn't matter that there's nothing at the DOCTYPE 
>level which has a URI, and that therefore there's no way of making 
>metadata statements about a DOCTYPE.

I personally have little use for the closed DOCTYPE idea, but I wouldn't 
want to deny to others any utility they may find there.  I understand that 
XML schema is intended to provide all the capabilities of DTDs, without 
imposing a single-language-structure on a document -- my vision of the 
future lies that way.

>[[Or do you believe that the DTD URI should retain a role of naming the 
>resource which is the DOCTYPE even once DTD based validation has become 
>obsolete?  If so, we should probably be having the forbid/absolute/literal 
>debate about DTD URIs too.]]

See above.

>You believe either (a) that the several HTML dialects are one language, 
>and that there is no need to be able to make different metadata statements 
>about them, or (b) that the single namespace decision was wrong.

(a) No.

(b) What "single namespace decision"?

#g

------------
Graham Klyne
(GK@ACM.ORG)

Received on Wednesday, 21 June 2000 12:17:36 UTC