This is a) a heuristic, b) very slow as it is synchronous, and c) wasteful (as we don't store the response but only check for the existence of a resource at the url). But it seems to work.