> The three results show that the encoding is being considered, but for
> us-ascii the output is broken.  What I'm expecting is that 
> the processor
> should be smart enough to use the � notation for extended ascii
> characters when in us-ascii mode.  Maybe I'm expecting too much.

I thought you already told us you had found documentation that explained
this behavior. Which doesn't make it any better. Sounds like you need a
different XSLT processor - fortunately there are plenty to choose from.

Michael Kay