RE: XML 1.1 Last Call Comments from the XSL WG

> Kay, Michael scripsit:
> 
> > Are you saying that normalization should be done when producing a 
> > final result tree from the intermediate result trees, or are you 
> > saying it should be done when serializing the final result tree?
> 
> Either, but I meant the latter.
> 
> The normalization constraints apply only to actual XML 
> documents used in
> interchange.   What a program does internally to itself is 
> not constrained.
> It is up to the XSL WG whether to require normalization 
> internally (but I personally recommend against it); what 
> matters is that when you generate an XML 1.1 document, you 
> ensure that it is normalized.
> 

I'm a little concerned that when you run two transformations in tandem, you
will get different results when the result of the first transformation is
passed directly to the second as a tree, from when it is serialized and
re-parsed. We have always had the principle that parse(serialize(tree)) is a
no-op. 

To me, normalization and serialization are separate operations and I think
it might be a mistake to couple them.

Michael Kay