Thanks, David and Michael.

On Apr 30, 2007, at 7:22 PM, Michael Kay wrote:
>Saxon doesn't know that the original document contained numeric character
>references - it only sees what the parser hands over, which is Unicode
>characters. You can force the characters to be output as numeric character
>references by using a character encoding in which they aren't present, for
>example <xsl:output encoding="iso-8859-1"/>, but they won't necessarily be
>exactly the same as in the input.

That makes sense.  Changing the encoding attribute on the xsl:output element looks like it did the trick (right now it's US-ASCII).

On Apr 30, 2007, at 6:44 PM, David Carlisle wrote:
>beware there never will be an XSLT 1.1, there was an early working draft
>but it was killed off to focus (eventually) on XSLT2. So to write
>standard conforming code you ought to move down to 1.0 or up to 2.0.

Ah, I had no idea!  I switched to 1.1 after the parser complained about not being able to access a result tree fragment stored in a variable using 1.0 and recommended using 1.1.  I'll make sure to use one of the two standard versions in the future and on this one since that part of the code has since been changed.

>Don't use d-o-e unless you know the consequences (I think you don't).
>Don't use d-o-e ever is a good rule!

After looking it up on google and catching the conversation on dpawson's site, I didn't understand this attribute well enough to know that it wouldn't work.  I think I get it now.

-Robert