[ghc-steering-committee] GHC2021 extensions should be sufficient for early intermediate Haskellers

[Richard Eisenberg]

Hi all,

In thinking about how I feel about TypeFamilies (and why I lean against inclusion), I realized I had a new criterion:

* GHC2021 extensions should be sufficient for early-intermediate Haskellers

Today, users have to enable a number of extensions just to get basic work done (I'm looking at you, FlexibleContexts). So they learn just to take whatever suggestion is presented to them in an error message and apply it.

But if extensions were considered more exotic, then users might not be so willing to add an extension. They would have to consider whether they really want to opt into more challenging error messages and a wider horizon. TypeFamilies is an extension that significantly widens the horizon, but also invites new scary error messages. I think it should be opt in. But this gatekeeping works only if users will be thoughtful about enabling the extension; thus my new criterion.

What do we think about this?

(I thought of putting this on Kialo, but it didn't seem to fit the setup there. Maybe I've erred in not just blasting ahead.)

Richard