an actual MS IE XSLT issue

Hi,

the discussion about who caused XSLT "to fail" reminded me of an 
*actual* XSLT issue I recently found with IE.

The problem seems to occur when serving XML with an XSLT PI, and the 
XSLT is running in HTML output mode, and is outputting unicode 
characters which the selected output encoding can't handle.

For instance, try 
<http://greenbytes.de/tech/webdav/draft-ietf-webdav-bind-24.xml>, which 
uses <http://greenbytes.de/tech/webdav/rfc2629.xslt>.

In IE, I'm getting a truncated result, and the error message

"Zeichen konnte nicht mit 'iso-8859-1'-Kodierung gespeichert werden"

And sure enough, when I change the XSLT's output element to use utf-8, 
the problem goes away. The problem also occurs with us-ascii.

Now, this really doesn't make sense. If the output encoding can not 
directly encode the character, it should just be encoded using a 
character reference.

Sure enough, this works as it should when running the transformation 
from the command line using msxsl.exe (a command line wrapper around 
MXSML's XSLT engine).

So far I haven't succeeded creating a minimal test case, but may be 
somebody else has seen this before?

BR, Julian