Re: [Haskell-cafe] Impact of "try" on Parsec performance

I'd suggest `try` can make parsers had to debug and fragile when refactoring. `try` is very useful for char parsers, where you don't really want to be sharing a prefix amongst different parsers. For regular parsers you have equivalents of the EBNF operators plus the `sepBy` etc. parsers so the "obsessive" left recursion removal you see in text books isn't so burdensome - you have combinators to help you do it rather than having to rely on introducing more productions.