Re: an xslt question re modeling namespaces

[james anderson <james.anderson@attglobal.net>]

David Carlisle wrote:
> 
> > Are these xslt-specific constraints? Do they derive from one of the data
> > models?
> 
> They are there to ensure that the tree corresponds to a tree that could
> have been derived by parsing a file in XML syntax.
> 
> If you copy element and attribute nodes from somewhere into the result
> tree you end up with a tree in which elements are in specified
> namespaces but there are no namespace nodes which correspond to any
> namespace declarations. The fixup just ensures that the system adds
> suitable namespace nodes.

I understand why the suggested "fixup" process may be useful in the
situations which the draft describes (that is, subtree output and/or
excision). I have not yet understood it to be necessary. The information
specified in the infoset suffices. Which does not include expanding all
the namespace nodes into the respective element node. Any missing
information can be constructed on the fly.

The question remains, where do the constraints on literal ns-node
presence arise?

> 
> In XSLT 1.0, this is be part of the output serialization requirements,
> namespace _declarations_ have to be added as the result tree is
> linearised, but in XSLT 1.1 draft you can query the constructed trees,
> so the "fixup" has been moved to an earlier point fixing up nodes in the
> tree, rather than just on linearised output (At least I assume that's
> the reason)
> 

My quandry is that the constraint has not only been moved to an earlier
point, it is specified as applying "a tree constructed by parsing an XML
document". 

? ...