I don't care about the politics, I just care about the data and what
works best. I want Theā„¢ good answer that works for newbies and experts
alike on the Haskell homepage. The HP was on the home page when I made
it, then everyone I spoke to told me they don't use HP, so I looked
around at what people used and replaced it with those things (MinGHC
effort, the OS X effort and hvr's prepackaged Ubuntu etc).

What's the regularity of the new release cycle going to be? I'm
preparing some survey questions and would like to put in:

* If the HP release cycle becomes <this>, will that change whether you
use and recommend it?
* The HP will be more in sync with GHC releases now, so that removes
one hinderance.
* I'll put in the survey the size of the download as a factor.
* Is the problem of upgrading beyond packages in HP still a thing,
such as the network package? I see this mentioned a bunch. Basing on
MinGHC should mitigate this, AIUI.
* Can you smoothly upgrade to newer packages generally?
* Are the requirements for authors still so stringent that authors
don't contribute? If so, the above question is important so that users
can still get those packages via cabal.
* Are we going with that "slimmed down" version that I seem to recall
mentioned in the last discussion that's basically a working GHC with
Cabal?

And any other questions where people were considering something a
dealbreaker that is now fixed in HP.

I'll also float by the set of questions here and on reddit before
actually publishing it. Then we can float it for around a week. I'd
like to be able to have either a definite answer that this will solve
the main gripes people have and everyone can move over to it, or
otherwise what's necessary to improve the situation, or whether HP +
alternatives will be the best possible solution.

I don't know that I'll publish a survey yet, but it seems like a good
idea to know what everybody thinks rather than going back and forth
about what we think they think. If we get this right it can dispel any
FUD about HP as a side effect.
_______________________________________________
haskell-infrastructure mailing list
[email protected]
http://community.galois.com/mailman/listinfo/haskell-infrastructure

Reply via email to