Henning Thielemann wrote:
I propose to remove the MonadFix instance for the ExceptT transformer, because it gives the illusion that it can handle exceptions, which it cannot.
I think this analysis is not quite correct. The exception is not handled but propagated outside the mfix, which is what I would expect: ghci> runExcept (mfix (\_ -> throwE ())) Left () The only case where one gets to see the artificial bottom introduced in the exception case, is when one incorporates the fixpoint value into an exception: ghci> runExcept (mfix (\v -> throwE v)) Left *** Exception: mfix ExceptT: Left That looks reasonable to me, because indeed no value is being produced by the inner monad action. Note that 'mfix' is a partial function to begin with; one always has to be careful about using the produced value inside the argument of 'mfix'. -1 from me on the proposal. Cheers, Bertram