Norman Walsh wrote:

>  / Jonathan Borden <jborden@[...].net> was heard to say:
>  | The problem is that for the internal subset, no mapping is needed
because
>  | the prefix->URI mapping can be obtained from the content context. The
> 
>  I don't think you can rely on the parser to have evaluated the xmlns
>  declarations on the root element before it reads the internal and
>  external subsets, so I don't think the distinction between internal
>  and external subsets is relevant.
> 
Let me clarify. The problem with using a prefix:name in a standalone DTD is
that nowhere within the DTD file is the prefix<-> namespace mapping defined
(i.e. akin to the "Prefix Defined" constraint). In an internal subset, the
prefix, though used within the DTD, is defined within the document content
(not necessarily just the root element).

It is not required that the parser have evaluated the xmlns declarations
prior to reading the root element, only that when actually validating QNames
be expanded.

Jonathan Borden
The Open Healthcare Group
http://www.openhealth.org