namespace rules were broken

[Lofton Henderson <lofton@rockynet.com>]

As I was editing, I discovered that the XCF namespace stuff was
broken.  Details:

The dtd said this:  xmlns CDATA #IMPLIED
Nothing else in the text required that the WebCGM namespace be defined in
XCF instances.  I believe that this violates the "XML
Namespace" standard.  (But I'm on slightly shaky ground with
this NS stuff.)

Two possible solutions:

Opt.1
=====
change dtd to:  xmlns CDATA #REQUIRED
This would mean that the XCF/WebCGM namespace would always have to be
defined on the 'webcgm' element, either as the default namespace (no
prefix) or prefixed.  As now, application-specific namespace always
must be defined as well.

Opt.2
=====
change dtd to:  xmlns CDATA #FIXED
http://www.cgmopen.org/schema/webcgm/
This would allow you to omit the xmlns attribute from the 'webcgm'
element for the WebCGM XCF namespace.  As now, application-specific
namespace always must be defined.

Further discussion
=====
Opt.2 has the side effect, *I think*, that you could never define another
namespace, other than XCF WebCGM, to be the default (no prefix)
namespace.  Is that correct?  I.e., this would be valid under
Opt.1 but not Opt.2:
<xcf:webcgm xmlns:xcf=
"http://www.cgmopen.org/schema/webcgm/"
            xmlns=
"http://example.org/asd/"
...>
[...companion file contents...]
</xcf:webcgm>

It looks to me like SVG11 does something approximately like
Opt.2.

Recommendation
=====
I have none now.  The way I left it is approximately Opt.1, because
it was easiest.  In the existing examples, I did add a xmlns
declaration, because that is never wrong.  (The question is about
*necessity*.)  

Thoughts?  (Or corrections to my analysis?)

-Lofton.