RE: Update on namespaces

At 09:25 1997 06 11 +0100, Martin Bryan wrote:
>If we use the notation concept for lexical typing of element contents (which
>I would like) how could we ensure portability of the specifications? If the
>rules are defined in something identified by a URL must that object return a
>Java applet that can be used to validate the contents of the element? If so,
>how can users control the applet? (For example, how can I say that I want
>the applet I just retrieved from the US to display the date in UK
>day/month/year format rather than its default month/day/year format?)

If we prefix element names with some "namespaceid:" thing, then we have
accomplished the syntactic "uniquification" of names in the document
instance without any mapping of the namespaceid bits to URIs.  By mapping
the namespaceid to URIs, we have now accomplished global uniquification.
That's all I think something like the XML spec would have to say.  All
other typing (lexical or otherwise) is an orthogonal issue that might
be addressed by some XML-data-typing spec, but that is true for both
the old fashioned single namespace DTD as well as the multiple namespace
problem, so let's not mix data typing issues with namespaces issues.

The URI to which a namespaceid is associated doesn't have to do or return
anything.  It doesn't even need to resolve to an actual physical thing
(though it may well make sense for it to resolve to some DTD or other
set of declarations)--it just provides a globally unique naming scheme.

Perhaps my suggestion to use a notation declaration to map a name to
a URI is confusing the issue.  Forget the fact that we all think that
notation has something to do with data typing--I don't believe such an
association is required by 8879's notation declaration--and just look
at the notation declaration as mapping a name that is local to an instance
into a structured name that is globally unique (i.e., a URI).  Consider
that "namespaceid:foo" just says that the name foo comes from the namespace
identified by the URI to which "namespaceid" is mapped and forget data
typing at this level. 

paul

Received on Wednesday, 11 June 1997 07:19:40 UTC