#12731: Generic type class has type family; leads to big dep_finsts -------------------------------------+------------------------------------- Reporter: ezyang | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.1 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by goldfire): What a great blog post to link to. (Although it did remind me of reading Neal Stephenson's novel ''Anathem'', which takes place in an alternate universe where just about all technical concepts have been renamed.) I will summarize it: The author proposes and does an in-depth comparison of several approaches that can be used to avoid incoherence of either type families or classes (it's the same problem) by doing local checks instead of a check at module-import-graph join points. All of these rules prohibit orphan instances of any kind, but are otherwise much subtler than I would have guessed. While we could explore this possibility (does anyone use orphan type family instances?), I see a simpler approach: treat `Generic` like we do `Typeable`, by requiring that GHC provide the instances. Does anyone hand- write `Generic` instances? If we know that GHC has written the instances, then we can be sure of coherence. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/12731#comment:1 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler