That last post reminded me of another issue I had seen in the past. I've done some pretty extensive work with JScript and programming IE as a fat client. In the process, we used XML and XSLT heavily.
One of the problems I've seen in the past is that the IE HTML DOM also mangles HTML. You can put well-formed XML into the HTML output, but you can't seem to get it out.
The IE DOM mangles it a somewhat predictable way, at least - removes quotes for all one-word attribute values, capitalizes the case on (most) elements, and removes closing tags for certain empty elements (like <br/> for example).
The only viable solution we've come up with was using XML data islands, which are also a pain in the ass because you have no control over the XML version - it just uses the default parser. If you want to do an XSLT transformation (as opposed to the bastardized XSL standard - proper XSLT support isn't there until 2.6), this can be a problem. You basically have to serialize the DOM and reload it to ensure the version - a performance problem to say the least.
Has anyone dealt with similar issues? If so, how did you work around them?