Notes from Ben's "contribute to ghc" discussion

Simon Peyton Jones simonpj at
Tue Sep 27 14:22:28 UTC 2016


Wow!  I didn't expect my scrappy notes to generate so much traffic!

Some quick thoughts:

* My notes were typed in real-time, during an open 70-person
  discussion during the Haskell Implementors Workshop.  Air-time was
  limited, and I just wanted to capture suggestions that people made.
  They do not reflect a thoughtful agreement or a concrete plan.

* As it happens, some of the subsequent traffic illustrates rather
  well the challenges that I wanted to address in my "Respect" post to, although I posted the latter before seeing the former.

  The good thing is that we ALL share the same goals: to make it as
  easy as possible for both newcomers and hard-core developers to
  contribute to GHC; to preserve GHC's conceptual integrity; to
  operate within the limits of what a bunch of volunteers can do.  We
  may differ in our judgements about how best to achieve those goals,
  but I'm certain that, if we take a little care, we can do so in the
  language of colleagues not adversaries.

  Happily, everything has calmed down a bit now, but still I'd like
  to renew my plea for courtesy; and in particular, to start from an
  assumption of good faith.  Nobody here is seeking to be hostile,
  dismissive, or excluding.  If my behaviour appears to you to be any
  of those things, please talk me privately, not in public; I have
  probably just failed to express myself well.

* Turning to the main issue of substance -- reducing the barrier to
  entry for new contributors -- one plea is "Just do it on Github, the
  same as everyone else".  I can see the force of that argument;
  Chris Allen calls it "legibility": simply being similar to other
  workflows reduces the barrier to entry.  (Chris and I had a useful
  conversation last night; thanks Chris.)

  I do not have a well-informed opinion about whether Github can do
  the job for us -- it's not the same Github as when we last
  consciously decided not to go that route.  Even if we stick with
  Phab we could probably do a better job of explaining the workflow,
  so that someone new is in no doubt about how to contribute.  But the
  choice of technology is, in the end, a judgement call about the
  balance of plusses and minuses.

* I really like Jakub Zalewski's suggestion of having a GHC-specific
  StackOverflow instance.  StackOverflow seems to have captured a
  great way for people to have a technical questions and answers.
  That might be better than the GHC wiki, or at least a great
  complement to it.  Better still, Jakub has volunteered to spin one
  up, an offer I think we should grab with both hands.

* I'm open to the idea of mentors -- if we could find enough people
  willing to act as mentors. I'm not confident we have enough supply
  to meet the demand, but perhaps we should try and see?
* It's worth remembering that we are in the midst of revising the
  process of how to propose a change to GHC, and the language it
  compiles, in direct response to feedback from the GHC developer

Onward and upward


More information about the ghc-devs mailing list