> To preempt the argument that ordering becomes an issue
> (namely, did the read occur before or after the
> write), I'll reiterate a sentiment posted in a thread
> to this forum.  Put a stake in the ground and say that
> all writes occur at the end of the transformation. 
> All reads see the documents prior to modification.  Is
> there a flaw in this reasoning that necessitated this
> restriction? 

Please correct me if I'm wrong, but I see no reason why the parser must
finish reading the input document before starting to write output.
I suspect (although I haven't read the code) that fast transformers
would do this.

	Ronan