Are you saying that it's not possible for a processor to
output in a
us-ascii document? I thought that was the whole point of the
notation -- that it can be represented entirely in us-ascii.
This is what
I've been getting at: is it reasonable to expect a processor
to leave
as precisely those characters when outputting in us-ascii. I
believe the
answer to be yes.
It's entirely reasonable to expect an XSLT serializer to output an NBSP
character to an ASCII output stream using one of the available escape
mechanisms. What you can't expect is to control which escape mechanism it
uses. Though Saxon lets you do that too.
Michael Kay