> BREAKS XPATH
> 
> (Paraphrasing) By allowing the XPath in an <assert> element 
> to reference ancestors, cousins, and siblings it will break XPath.

I don't recognize or recall the comment that you claim to be paraphrasing.
> 
> POOR PERFORMANCE
> 
> As I illustrate above, in XSLT the XPath expressions can 
> reference anywhere. It doesn't seem to impose an undue burden 
> on XSLT processors. So why would it impose an undue burden on 
> XML Schema validators?

I think this is based on an assumption that the performance requirement is
higher for validation than for transformation. One could question whether
this assumption is well-founded - it's obviously going to vary from one user
to another.
 
> Besides, even if it did, so what? Isn't the objective to make 
> things easier and better for the users, not the validator 
> implementors?

Well, it's users who benefit from improved performance, not implementors!
But I agree entirely. My argument for this feature is not based on
performance.

Michael Kay
http://www.saxonica.com/