Well-balanced fragments as input to XSLT 2.0?

I would like to propose that XSLT 2.0 should be able to accept the same
kind of "well-balanced" fragments as it can output.

The exact details of what it can output are slightly complex (see The
XSLT Tree Model and XSL:Output sections in Mike Kay's XSLT Programmer's
Reference for a good explanation) but the basic attraction is that it
includes options of outputting XML with no root element (pure text or
CDATA), one root element (the traditional well-formed XML) or multiple
root elements (an XML hedge?), though in all cases every open tag must
have a close tag in the usual way.

For me, this is great. It really adds to the utility of that otherwise
slightly confusing distinction between the root element and the root
node (I know, the root node can also hold root comments and PIs) and I
think it would add to the power and simplicity of XSLT if you could
apply XSLT to any well-balanced XML fragment that you could acquire.

I see one problem - you can create these well-balanced fragments
programmatically, through DOM operations (or through XSLT returning DOM
objects) but you may not be able to parse them directly with XML parsers
since they may not be well-formed. People may find this discrepency
disturbing, but I find this to be outweighed by the attraction of
achieving symmetry between XSLT's input and output models.

Any other pros or cons?

Francis.

 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list