[Due to an upgrade in Yale's mail system, this message (and a few others) didn't get through to the Yale half of the Haskell mailing list - so I'm resending. I guess both halves of the list have already seen Kevin Hammond's response that it's not true... so it goes. -- Alastair] Chris Dornan ([EMAIL PROTECTED]) asks: > Has anybody considered limiting the 1.3 upgrade to the new > much-improved I/O libraries? Several committee members who'd love to do just that. My (personal) view on why we're not limiting the upgrade to just the I/O libraries is that (most of) the other new features are necessary to properly support the I/O libraries. * The I/O monad seemed to be on the point of stealing the names >>=, >> and return for the exclusive use of the I/O monad. Adding constructor classes avoided that. * The LibTime module will benefit significantly from the addition of records (and may well benefit from the addition of strictness annotations). We could omit LibTime from Haskell 1.3 and produce Haskell 1.4 sometime next year with records and LibTime - but we'd rather get all the changes over with in one go. We also felt unhappy about encouraging programmers to use monads without providing some syntactic sugar to make them palatable. (Old time monad hackers have become used to the "gzinto"-style of programming that goes with it - but it's not much fun for beginning programmers or those who have to teach it.) We could omit newtype at this stage but we chose to add it in anticipation of defining further standard libraries. The current design makes good use of newtype to preserve abstraction within implementations and to enable the use of type/ constructor classes (you can't define an instance for a type synonym). We could also omit other changes (deletion of n+k patterns, various prelude changes, ...) but I don't think it would speed the process up. It might break less code - we'd certainly welcome comments on which changes will break a lot of code (especially if the benefit from changing it seems insignificant). Alastair Reid Yale Haskell Project
