importhtml5libstring=u"<p>name&thinsp;: value</p>"dom=html5lib.parse(string,treebuilder="dom")printlen(dom.getElementsByTagName("p")[0].childNodes)

This gives three, split around the\u2009. Per spec, we should have one child text node, I'm pretty sure. This only shows up here because etree/lxml can't have adjacent text nodes in their data model.