I'm just saying: let's _not_ add ExistentialQuantification in GHC2021 (because it ought, in my opinion, to be considered deprecated). Let's not go into talking about chopping the feature into smaller pieces. I do agree that one of my arguments in the original email has been shown to be incorrect, though. But I still think that my suggestion holds.

On Fri, Dec 4, 2020 at 3:34 PM Richard Eisenberg <rae@richarde.dev> wrote:

I do use the data Ex = forall a. ... syntax occasionally, but only when I'm emphasizing that I'm not building a GADT. I would entertain the thought of removing it. Yet, I don't think we can do so as part of this process, as we don't have a combination of extensions that allows GADT-syntax existentials but not this other syntax. (Even -XGADTs -XNoExistentialQuantification doesn't eliminate it.)

In any case, I've voted against GADTs, and so if you want to disentangle this one piece of syntax, that would also be a vote against GADTs, followed up with a proposal to spin out this syntax into its own extension.

Richard

On Dec 4, 2020, at 9:14 AM, Spiwack, Arnaud <arnaud.spiwack@tweag.io> wrote:

Ah one clarification, GADTSyntax alone does not enable existential types. You need to enable either ExistentialQuantification or full GADTs.

It makes more than perfect sense, but I wasn't aware of this. A consequence of always enabling -XGADTs, I suppose :-) .
_______________________________________________
ghc-steering-committee mailing list
ghc-steering-committee@haskell.org
https://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-steering-committee