XPointers without requiring xmlns()

> I think that the major argument against the requirement for scheme names
> as QNames are the ease-of-use and aesthetic issues that it imposes on
> schemes defined by organizations other than the W3C.
> I have a proposal that would eliminate the necessity for xmlns() prefix
> definitions to be included as part of an XPointer for a  processor.  It is
> in fact a rather small change to the existing specification:
> Rather than requiring of an XPointer processor that: "The initial
> namespace binding context prior to evaluation of the first pointer part
> consists of a single entry: the xml prefix bound to the namespace name
> http://www.w3.org/XML/1998/namespace. "
> Instead mandate that: "The initial namespace binding context prior to
> evaluation of the first pointer part consists of at least a single entry:
> the xml prefix bound to the namespace name
> http://www.w3.org/XML/1998/namespace.  Additional namespace bindings may
> be configured in the initial context in an implementation specified
> manner."
> That way, Simon's ugly example no longer requires the "xmlns()" part given
> an XPointer implementation that supports configuration of additional
> namespaces.  If someone wanted to use his scheme in a fashion that would
> be compatible across a wider variety of XPointer implementations, then
> they could still include the xmlns() ugliness.  Since using xmlns() in the
> pointer would automatically override any existing definition, there is no
> "compatibility" issue.
> This proposal eliminates the ugliness with a rather simple change to the
> XPointer framework specification, leaving the standardization of the
> implemention to some group that doesn't expire in the current month, yet
> still provides for the same guarantees of uniqueness, to the point of even
> retaining the use of namespaces.
> 	Keith
>