View source for DOM/ContentIterator Issues

Content Iterator, we've discovered, has a design flaw in it responsible for [https://bugzilla.mozilla.org/show_bug.cgi?id=157373 several] [https://bugzilla.mozilla.org/show_bug.cgi?id=302775 bugs] in Mozilla.  In essence, nsContentIterator and its friends assume (incorrectly) a nsIContent interface is available for every nsIDOMNode object they receive.

The truth of the matter is, there are four non-content DOM node types (possibly five, if you count XPathResult):

* Attr
* Document
* <s>DocumentFragment</s> (nsDocumentFragment QI's to nsIContent)
* EntityReference

According to [http://www.w3.org/TR/2004/REC-DOM-Level-3-Core-20040407/core.html DOM Level 3 Core], all four of these node types can have child nodes.  These child nodes do implement nsIContent (except for EntityReference, which can itself be a child node of any of these).