Hi,
The reason I'm willing is because I believe that the community really wants it -- but I'd love to have evidence for that belief. This discussion is taking place on the GHC committee mailing list to which the community more broadly can't easily contribute. I'd love to ask them more explicitly, or even take a poll.
Well, let’s see :-) https://www.reddit.com/r/haskell/comments/je1t82/does_the_idea_of_xghc2021_e... Am Montag, den 19.10.2020, 15:21 +0200 schrieb Alejandro Serrano Mena:
Following Simon's ideas, should we maybe create a poll and some scripts to scrape Hackage/Stackage and inform our decision? For the poll I'm thinking on something like categorizing how much people "want" an extension to be on automatically, ranging from "yes, please!" to "never!". We could even gather a few more opinions about why the choice was made and selecting things like "stability" and so on.
I was pondering both a pool, and gathering statistics. I like to keep things lean, but it doesn’t hurt. And I think it is compatible with the process that I proposed: Of course we can put out such a poll at appropriate time, and let the community vote, and let every committee member take the outcome of that vote into account. I would prefer to _not_ make it a formal requirement (e.g. “at least 80% in the popular vote”), if only to avoid the discussion of how reliable online votes are. As for hackage statistics, I also think that gathering that data is useful. I don't actually think widespread use is such an important criteria, if an extension is either very narrow in the target audience, or is safe and convenient, but often not worth bothering writing out the {-# LANGUAGE … #-}. For example NumericUnderscores – a nice feature if you can use it without an explicit langauge extension, and lowering the barrier to the use of such polish is (I believe) one aim of GHC202x. But then expecting widespread use of the extension seems to be self-contradictory. So yes: Let us collect these stats, and let us take them into account. Let’s not make it a formal requirement, and an informal requirement only as “please take into account” (and then I can say that NumericUnderscores, for me, does not require a huge up-take before I would vote it in.)
Finally, I'm not at all tied to having a yearly process. If you think 2 or 3 years would be enough, that's 100% fine with me. In fact, my focus is to get the first GHC2020 out. But I think that telling people that there are new chances in a definite period of time helps in giving a mid-term perspective: we don't have to have the perfect set now, we can start and keep refining in the upcoming years.
Right, let’s _not_ argue too much about the cadence at this point. Let’s do the first one, and revisit again next summer, and maybe also listen to how it was received. Cheers, Joachim -- Joachim Breitner mail@joachim-breitner.de http://www.joachim-breitner.de/