Re: locally scoped element decls and namespaces

Henry S. Thompson wrote:
> 
> james anderson <James.Anderson@mecomnet.de> writes:
> 
> >
> > Wait. Is this really saying that one is looking for a way to encode the
> > specification, that a symbol, so encoded as to denote that it is in no
> > namespace, is really to be taken to be identical with another symbol, so
> > encoded as to denote that it is in another, definite namespace?
> > Isn't this impossible by definition?
> 
> I think you've misunderstood my (admittedly obscure) point.

No, I think that I understood you. I still understand you. The point is not at
all obscure. It turns out to be one of the disagreeable aspects of the
namespace spec should one try to implement a processor with a complete model
for names.

I could have either followed the spec strictly and left such names uninterned,
or I could have interned them in some "special" namespace. Given the
premonition, that situations such this would be bound to arise, I opted for
the second. As the spec says little about how names are to be processed, one
might say that I had little to go on. Much as you predict, I did what i
suppose was intended but I would note that this option is, strictly speaking,
in contradiction with the little which spec does say. Should it be what the
spec intends, then the wording should be appropriately revised to recognize
existance of a specially designated default namespace.

> Suppose I
> have an XML instance document with no namespace declarations of any
> kind governing some or all of the GIs therein.  These GIs are not in
> any namespace at all.  The Namespace REC even gives us a way of
> EXPLICITLY indicating that names in a certain scope are in no
> namespace, namely "xmlns=''".  But I may perfectly well want to
> identify a schema to use for schema-validating such an instance
> document.  How do I do so?

Whatever notational tricks you have up your sleve, the question remains, how
do you take these names which are "not in any namespace at all" and postfacto
put them in one? I mean, if they're not in one, how do you find them to
operate on them?

>  xsi:schemaLocation uses PAIRS of URIs to
> accomplish this, with the first member of each pair being a namespace
> URI.  But _ex hypothesi_ the GIs of the elements in question are not
> in a namespace, hence there is no namespace URI I can use in
> xsi:schemaLocation.  If we stay with the admittedly rather clunky
> syntax of xsi:schemaLocation as it stands, I expect we'll nominate
> something such as ##noNamespace for this purpose. . .

Thus the notion of the specially designated namespace. Sneaking it in the back
door is a bad idea.

Received on Wednesday, 5 January 2000 04:04:32 UTC