On 10/07/2011 08:51, Yitzchak Gale wrote:
Simon Marlow wrote:
In response to this ticket: http://hackage.haskell.org/trac/ghc/ticket/5275 ... === Proposed API changes === 1. DEPRECATE mkTyCon... 2. Add mkTyCon3 :: String -> String -> String -> TyCon ...Most users can just derive Typeable, there's no need to use mkTyCon3. In due course we can rename mkTyCon3 back to mkTyCon.
Anecdotally, it seems to me that many, if not most, packages on Hackage that create Typeable instances do so using mkTyCon, not by deriving Typeable. There seems to be some standard boilerplate that spreads virally from one library to the next.
Perhaps it would be worth checking to what extent mkTyCon is used on Hackage. If this is as widespread as I suspect, even a simple deprecation warning could have a cascading effect on typical output from cabal install.
If there are a lot of packages using mkTyCon (I've already encountered one, so you might well be right), what do you suggest? Deprecation is the gentlest way we have to signal that something needs to be fixed. Perhaps deprecation warnings should be piped into "sendmail <package-author>" instead of being printed :-) Cheers, Simon