On this one, I agree with David and disagree with James. I don't
see the advantages to using an exception. I think that a SAX processor
should use fatal() (why the longer fatalError()?) - this has the
advantage that you can, after the first message, go on looking for
more fatal errors. Of course, a SAX processor must not, after the first
fatal() callback, emit any more element() or charData() callbacks.
I also think we should add a lineOffset argument, as someone
suggested; but I don't think that SAX should *require* processors
to do entity/line/offset tracking, since it's hard; SAX should
specify that a value of -1 for any of these arguments means the
processor doesn't know. -Tim