Faltings proof *does* provide (with a bit of extra work) a bound on the number of exceptions (as does Roth's proof of his theorem), although not a bound on the size of the exceptions. (I mean a bound effective in the data). -Lou van den Dries-