I've got a lot of expereince with XML and XSLT and there's one thing I can say for sure: XSLT is slow, and it'd be even slower on a 200MHz processor. Also it's a lot of work to write an XML and XSLT processor so what I propose is that the kernel can either output HTML in the form it is now, or XML using ?.xml as the extension.

XML files can contain a PI (processing Instruction) that tells the parser ( in this case the browser) what XSLT file to use to transform the document.

This way we can use the existing HTML code that Mark has written, and just add some XML support and then we can write the XSLT ourselves and not burden Mark with it.

All of the processing would be done on the client machine so the empeg unit only has to dish out 2 files (+ any images and linked docs)

Just my 2 cents.

-B