Client-side XSLT-to-HTML transformation in Firefox in combination with custom namespaces might generate a mixture of uppercase and lowercase tag names (clearly a Firefox bug).

The data structures generated by Firefox after this transformation indicate to jQuery that it's working on a an XML document, but due to the mixed-case tags (in combination with isXML being true), the Sizzle filters don't work correctly.

Furthermore, the "getElementsByClassName" function is available in XSLT-transformed HTML document, so the Sizzle engine could use the efficient version of class matching algorithm. However, the Sizzle function using getElementsByClassName returns immediately (again due to isXML flag = true).

To work around the Firefox bug, the isXML and isXMLDoc functions should identify the XSLT-transformed document with HTML root node as HTML, not XML. But because FireFox retains lowercase tag names in the XSLT transformation results (regardless of whether the xsl:output is set to HTML or XML), the root element node name is not "HTML" but "html" and the isXML functions return "true".

To fix this bug, we simply have to use the toUpperCase() function when testing the root element name:

var isXML = function(elem){

return elem.nodeType === 9 && elem.documentElement.nodeName.toUpperCase() !== "HTML" ||

!!elem.ownerDocument && isXML( elem.ownerDocument );

};