I think the parsing algorithm should take the prompt="" attribute of <isindex> in account. It replaces the string of characters placed before the <input> element with its contents. (In that case there will be no characters after the <isindex> element.)
Also, the prompt="" attribute will not be on the <input> element afterwards.
-- Anne van Kesteren <http://annevankesteren.nl/> <http://www.opera.com/>
