[ghc-steering-committee] #380 GHC2021: How to proceed?

[Joachim Breitner]

Hi,

Am Montag, den 18.01.2021, 10:32 +0000 schrieb Simon Peyton Jones via
ghc-steering-committee:
> Can you share the final proposal with us? Currently it says "Proposed
> change: TBD" which isn't quite the final version. '

Updated:
https://github.com/ghc-proposals/ghc-proposals/blob/ghc2021/proposals/0000-ghc2021.rst#motivation


>   Also the proposal doesn't have a motivation section.  Perhaps we
> should ensure that the proposal stands by itself.

I added two paragraphs, but mostly pointing to the motivation for
GHC22xx in general, and adding one mentioning the particular
considerations for the _first_ of these.

I feel like that suffices, but if you’d like to elaborate, feel free to
edit that PR, or suggest changes on 
https://github.com/ghc-proposals/ghc-proposals/pull/380/files

exclude extensions like OverlappingInstances, which may be stable but
we don't want to be on-by-default.

> [in the user manual too]
> Or some summary anyway, so that users know what the intended use of the extension is.

Sure, the actual implementation MR to GHC can give more guidance, but
usually that doesn’t need to be written down in detail in the GHc
proposal?

> Is there `-XNoGHC2021`, and if so what does it do?  Proposal should say.

The meta-proposal says

   This language version can be used as a language extension (with the
   sole effect of implying other language extensions), but also as a
   language versions in places where Haskell98 or Haskell2010 is valid
   (e.g. via Cabal’s default-language field).

   When ghc is used without an explicit language choice, the latest
   GHC20xx known to GHC is used. This applies in particular to uses of
   ghci.

and I’d hope that this answers such technical details (and that the GHC
devs will resolve obscure corner cases sensibly)?

It seems that ghc-8.8 does not know -XNoHaskell2010, so I would assume
that there is no -XNoHaskell2010 either. If there is a nice use case
for that, I think adding this a feature doesn't need committee
involvement, and can be done by GHC contributers.

Relatedly, at least one community member asked for a way to opt-into
the current “default” mode of GHC (which is neither Haskell98 nor
Haskell2010, e.g. it has NondecreasingIndentation).
He suggested simply calling that GHC2020. Whether that happens, and if
it’s called GHC2020 or OldDefault or something else, I’d also trust the
GHC devs.


Cheers,
Joachim


-- 
Joachim Breitner
  mail at joachim-breitner.de
  http://www.joachim-breitner.de/