computed namespace nodes

As I read the current specification, the semantics of computed namespace
nodes seem a little peculiar.  The ability to bind namespace URIs at runtime
poses some problems:  suppose one has a computed namespace node (which binds
some URI, only known at runtime, to a fixed prefix), and one then calls a
function within the scope of this namespace declaration, which makes use of
the prefix.  How is one to know during the static analysis phase which
function is meant, given that the URI of the function name is not known
until runtime?

 

Am I reading this wrong?

 

Damien