How, precisely, can we improve?

Simon Peyton Jones simonpj at
Wed Sep 28 08:36:33 UTC 2016

Interesting article.  Michael suggests using markdown in repo-controlled files rather than a wiki.  I can see the force of that. Maybe we should consider it.


From: Alan & Kim Zimmerman [mailto:alan.zimm at]
Sent: 27 September 2016 15:54
To: Simon Peyton Jones <simonpj at>
Cc: Sven Panne <svenpanne at>; ghc-devs <ghc-devs at>
Subject: Re: How, precisely, can we improve?

I think this is relevant to the dicussion:<>

On Tue, Sep 27, 2016 at 4:22 PM, Simon Peyton Jones via ghc-devs <ghc-devs at<mailto:ghc-devs at>> wrote:
We currently have *3* wikis:<>

I didn’t even know about  the third of these, but the first two have clearly differentiated goals:

•<> is about user-facing, and often user-generated, documentation.  Guidance about improving performance, programming idioms, tutorials etc.

• is about GHC’s implementation, oriented to people who want to understand how GHC works, and how to modify it.

I think this separation is actually quite helpful.

I agree with what you and others say about the difficulty of keeping wikis organised. But that’s not primarily a technology issue: there is a genuinely difficult challenge here.  How do you build and maintain up-to-date, navigable, well-organised information about a large, complex, and rapidly changing artefact like GHC?  A wiki is one approach that has the merit that anyone can improve it; control is not centralised.  But I’d love there to be other, better solutions.


From: ghc-devs [mailto:ghc-devs-bounces at<mailto:ghc-devs-bounces at>] On Behalf Of Sven Panne
Sent: 27 September 2016 08:46
To: ghc-devs <ghc-devs at<mailto:ghc-devs at>>
Subject: Re: How, precisely, can we improve?

Just a remark from my side: The documentation/tooling landscape is a bit more fragmented than it needs to be IMHO. More concretely:

   * We currently have *3* wikis:<>

     It's clear to me that they have different emphases and different origins, but in the end this results in valuable information being scattered around. Wikis in general are already quite hard to navigate (due to their inherent chaotic "structure"), so having 3 of them makes things even worse. It would be great to have *the* single Haskell Wiki directly on<> in an easily reachable place.

   * To be an active Haskell community member, you need quite a few different logins: Some for the Wikis mentioned above, one for Hackage, another one for Phabricator, perhaps an SSH key here and there... Phabricator is a notable exception: It accepts your GitHub/Google+/... logins. It would be great if the other parts of the Haskell ecosystem accepted those kinds of logins, too.

   *<> has great stuff on it, but its relationship to<> is unclear to me. Their "documentation" sub-pages look extremely similar, but<> has various (great!) tutorials and a nice overview of common libraries on it. From an external POV it seems to me that<> should be seamlessly integrated into<>, i.e. merged into it. Having an endless sea of links on<> is not the same as having content nicely integrated into it, sorted by topic, etc.

All those points are not show-stoppers for people trying to be more active in the Haskell community, but nevertheless they make things harder than they need to be, so I fear we lose people quite early. To draw an analogy: As probably everybody who actively monitors their web shop/customer site knows, even seemlingy small things moves customers totally away from your site. One unclear payment form? The vast majority of your potential customers aborts the purchase immediately and forever. One confusing interstitial web page? Say goodbye to lots of people. One hard-to-find button/link? A forced login/new account? => Commercial disaster, etc. etc.

Furthermore, I'm quite aware of the technical/social difficulties of my proposals, but that shouldn't let us stop trying to improve...


ghc-devs mailing list
ghc-devs at<mailto:ghc-devs at><>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the ghc-devs mailing list