Doru's suggestion of PP + islands is good and makes building complex parsers easy, but his toy example gives wrong results: <script> function end() { document.writeln('</script>'); } function start() { document.writeln('<script>'); } </script> or: <script> alert('hello'); // <script> alert('world') </script> or: <!--<script>alert('hello')</script>--> better to use one of the proper HTML parsers, at least if you need more correctness.