[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: Namespaces, schemas, Simon's filters.
- From: "Fuchs, Matthew" <firstname.lastname@example.org>
- To: 'Jeff Lowery' <email@example.com>, firstname.lastname@example.org
- Date: Wed, 29 Aug 2001 15:26:55 -0700
> -----Original Message-----
> From: Jeff Lowery [mailto:email@example.com]
> Sent: Wednesday, August 29, 2001 2:47 PM
> To: Fuchs, Matthew; 'Peter Piatko'; firstname.lastname@example.org
> Subject: RE: Namespaces, schemas, Simon's filters.
> > It would be nice to coalesce
> > xsi:type and namespacing.
> Doesn't that tie namespaces to context? I though this was a Bad Thing:
No, in that the namespace specified in the _instance_ would not require
having the schema around. The instance would fully specify the namespace
(which might happen to be "created" by a complexType).
> > > Option (b) implies that an identifier might map to
> multiple types. I
> > > the question boils down whether this ok or not.
> > In the end, I think you have to allow it. My gut tells me that the
> > alternative could get quite nasty. Besides, one of the nice
> things about
> > XML is that it does let you do your own thing, even if that
> thing isn't
> > always the best thing to do.
> It is possible now to create two Schemas with the same
> targetNamespace that
> attach the same global elements to different types (you just
> don't want to
> use them in tandem). I think that is insanity, but nothing prevents or
> admonishes against it as far as I know.
Yes, just as you can create two java classes (or C++ classes) with the same
name but everything else different. Loaders will scream at you if you try
to use both simultaneously. Likewise, you can't use both Schemas in the
same "validation episode".
> Are identifiers and semantics orthogonal or not? Certainly the English
> language supports the concept of one word meaning multiple
> things, but it
> forces interpretation in context, which machines have a
> notoriously hard
> time with. Such forced interpretation seems contrary to the goals of a
> schema language and type enforcement system.
> I see nothing wrong with a rule that says if you're going to
> give something
> a name, give a unique name. There's not way to enforce that,
> of course, but
> we're dealing with it for namespace identifiers already.
> -- Jeff