This strips out HREF, VALUE and SRC attributes and comment nodes to be able to serialize the DOM and compare mostly the structure. This helps because HREF, VALUE and SRC tends to contain session identifiers with random data that give false positives when trying to diff the HTML code.

sanitize_dom.js

(function(){
	function removeAttr(attr){
		var xpElms = document.evaluate('//*[@'+attr+']', document.documentElement, null, XPathResult.UNORDERED_NODE_SNAPSHOT_TYPE	, null );
		var elm;
		for(var i=0; elm = xpElms.snapshotItem(i); i++){
			elm.removeAttribute(attr)
		}
	}
	removeAttr('href');
	removeAttr('src');
	removeAttr('value');
	// remove <!-- comment --> and text nodes
	var xpElms = document.evaluate('//comment()|//text()', document.documentElement, null, XPathResult.UNORDERED_NODE_SNAPSHOT_TYPE	, null );
		for(var i=0; elm = xpElms.snapshotItem(i); i++){
			if(!(elm.parentElement.tagName in {'SCRIPT':1,'STYLE':1}))
				elm.parentElement.removeChild(elm)
		}
})();