I agree, the specification seems to set authors up for failure. I never understood why the language needed & followed by WHITESPACE to be escaped. It seems like it would be pretty easy to ignore this sequence in a parser?