> Handling SML blocks seems appreciably harder to me, because you > actually have to tokenize the SML. Handling comments just requires > looking for "(*" and "*)". And knowing when you're inside a string, since then you don't want to count "(*" or "*)".