RE: computed namespace constructors

Thanks for the comment. In fact this problem with computed namespace
constructors has already been noted by the working group. The WG considered
two alternatives, making the effect entirely static (an equivalent to
xmlns="x" for use with computed element constructors) and making it totally
dynamic. The current spec which has both static and dynamic effects is
unworkable. The decision was made to go for a purely static approach - which
means that both the prefix and the uri must be known statically. It was
decided not to offer any facility for dynamic construction of namespace
nodes (analagous to XSLT's <xsl:namespace> instruction) in XQuery 1.0, on
the grounds that it's not needed very often. Unfortunately this decision was
made too late for the November draft.

Michael Kay

> -----Original Message-----
> From: Per Bothner [mailto:per@bothner.com] 
> Sent: 25 November 2003 08:01
> To: public-qt-comments@w3.org
> Subject: computed namespace constructors
> 
> 
> 
> The semantics of computed namespace constructors is unclear:
> 
>    3.7.3.7 Computed Namespace Constructors
>    A constructed namespace node ... binds a namespace prefix 
> represented
>    as NCName in the syntax) to a URI and adds the namespace 
> prefix to the
>    in-scope namespaces for its enclosing element.
> 
> But the "in-scope namespaces" is part of the *static* 
> context, and constains both prefix *and* URIs:
> 
>    [Definition: In-scope namespaces. This is a set of (prefix, URI)
>    pairs. The in-scope namespaces are used for resolving prefixes used
>    in QNames within the expression.]
> 
> So how can a run-time namespace expression add to the static 
> in-space namespaces?
> 
> It might be possible to change the sementics so only the namespace 
> prefixes is part of the static context, while the namespace 
> uri is part of the dynamic context, but that would certainly 
> complicate implementation and probably the semantics  Consider:
> 
> element {$ename} {
>     namespace xx { myns:f($myns:x) },
>     namespace myns
>        {if $version eq 1 then "http://example.org/v1"
>         else "http://example.org/v2"},
>     attribute xx:type {"myns:invoice"},
>     $content
> }
> 
> This has a forward reference to the myns namespace (is that 
> allowed?), and an function name, a variable name, and an 
> attribute name, all of whose namespaces have to be calculated 
> at run-time.  I assume the intention is to not allow that.
> 
> I suggest a separate concept of "active namespaces" 
> consisting of anmespaces declared in namespace attributes and 
> namespace constructors. This could be part of the dynamic 
> environment, but not the static environment.
> -- 
> 	--Per Bothner
> per@bothner.com   http://per.bothner.com/
> 
> 
>