Proposal: Add-on packs

[Jason Dagit]

On Wed, May 29, 2013 at 5:51 PM, Mark Lentczner
<mark.lentczner at gmail.com> wrote:
> On Wed, May 29, 2013 at 4:59 PM, Jason Dagit <dagitj at gmail.com> wrote:
>>
>> I expect it would be different in the way that anything installed as
>> part of the platform is different: It's a blessed version and part of
>> a coherent set of packages.
>
>
> But as an add-on pack, why not just publish WxHaskell-Stable with a set of
> .cabal constraints that tie it to a particular HP version? There really
> isn't anything "blessed" by being in the HP - it is just a stable set that
> the individual package maintainers pick. No reason the WxHaskell maintainers
> can't do the same.

If the wxHaskell team created a 3rd party HP add-on that might work.
It would be even better if the HP acknowledged that add-on and steered
beginners towards it.

Trying to respond to this, I run smack into the question of, "Well,
what value does/should the HP provide?" For me, the HP could provide
just ghc and cabal-install and I'd be happy. Then again, I don't
consider myself part of the target audience. I think people who want
to get started with Haskell, and have something simple but full
featured to install, are the audience. Would you agree?

My impression is that the primary value of the HP is in having the HP
sign off on things so that newcomers have some clue as to what
packages to get started with. The secondary value being the convenient
packaging itself.

> I've heard this argument before for inclusion of something in the Platform -
> and I don't understand it. The platform has no secret sauce for making
> things installable. If the Gtk2hs team can't find a way to make installing
> Gtk2hs easy, the Platform team certainly isn't going to be able to figure it
> out!

It's about reducing duplication and making things convenient for the
masses (the secondary value of the HP as I put it above). If the HP
team can get, say, Gtk2hs built in a redistributable way, then (the
logic goes) it would be nice to reuse that effort instead of forcing
users to reproduce a non-trivial to configure build environment. This
probably doesn't apply to linux where software tends to not be very
redistributable. In some ways it's like the difference between gentoo
and debian. There are pros/cons to both approaches (gentoo vs.
debian), and for some (most?) users being able to install a prebuilt
binary is a huge win.

Jason