The project description states that: All syntax errors detected in invalid ml programs must be reported via stderr on a line commencing with the '!' character. Your runml program must be able to detect all invalid ml programs - EXCEPT that your program will not be tested with any invalid expressions, so you do not need to validate the syntax of expressions. Does this mean that you do not need to validate any identifiers as an expression could be an identifier from the syntax outline posted: expression: term [ ("+" | "-") expression ] term: factor [ ("*" | "/") term ] factor: realconstant | identifier | functioncall | "(" expression ")"