> Hmmm.  I was afraid of that.  I am still baffled as to how to 
> go about telling my stylesheet that the input it gets from a 
> particular source tree by way of the document() function that 
> it will have already been escaped and therefore that '&' 
> need not be escaped again (making it '&').
> 

The document() function invokes an XML parser and it can only do what an XML
parser does.

In fact an XML parser removes one level of escaping, and a serializer adds
it back. So the parser turns "&" into "&" and "&amp"; into "&",
and the serializer turns them back into "&" and "&"
respectively, unless d-o-e is set, in which case they are turned into "&"
and "&" respectively. All the evidence is that your XML source as read
by the parser was actually double-escaped. This quite often happens when you
have fragments of XML stored in a database: if you try to extract it as XML,
and the database software doesn't realise that it's already in XML format,
then the database software adds a level of escaping that you don't want. The
way to get rid of it is to change the way you do the database query.

Michael Kay