HP comments from the State of Haskell survey
Don Stewart
dons at galois.com
Mon Aug 9 16:30:29 EDT 2010
The following comments about the Haskell Platform appeared in the
State of Haskell survey.
http://www.johantibell.com/files/state-of-haskell-2010.html
* Apart from then Haskell Platform, software installation is a pain.
Cabal install is bearable, if painful, for developers, but it is
unacceptable for end users.
* Haskell's libraries are quite good, but I think the language could
really stand to have a better core API. This is something that I think
Java did very well. The Java API provides a huge number of important
libraries including GUI support. Since it's shipped with Java and the
defacto standard it provides a much more reliable platform to build on.
I realize the Haskell Platform is our attempt to solve this problem.
However, because it's built on hackage, which has all the packages
anyway, you don't really feel the distinction between official Haskell
Platform packages and others that are less accepted. Also the issue of
manpower for engineering still somewhat limits Haskell in how well it
can compete with the Java API.
* If it counts though, I have had some problems with installing certain
packages on Cabal, but every time I use Cabal it seems like the problems
reduce in number, so that's a good sign. I use Arch Linux, so I ended up
just installing the Haskell Platform from the AUR, and now everything
works, so I guess it's a non-issue.
* The library situation is a bit muddled, but I have faith in the Haskell
Platform.
* Lack of a single dictator or vested organisation in charge of the
platform libraries. A large libraries/batteries included suite need an
coherent and consistent design that can only really come from a single
vision worked on by many as opposed to many visions each worked on by a
few.
* Consistent, documented, quality libraries -- the quantity is there, and
the Haskell Platform looks like the right vehicle -- we just have to get
more libraries polished and vetted.
* fragmented community and libraries. For example, when a new comer asked
me "what libs do I use to parse XML? JSON? Connecting to databases?
Create a simple web application?" The answer I always end up giving is
"Depends". Haskell platform solves this problem a little bit, but it's
only the first step.
* I also think Haskell platform isn't getting enough push. People still
think to try Haskell means install ghc. It doesn't help that both
haskell platform and ghc in various linux distro as well as MacPorts
gets lagged behind badly.
* Of course, the Haskell Platform could be a partial fix for that by
blessing certain libraries over others, but its progress seems very
gradual. Libraries developed separately may need to be merged before a
suitable "community-blessed" version can be included in the Platform.
But I worry that this effoert will stall due to an overemphasis on the
quest to find the perfect API.
* Packaging system for libraries and version dependencies. I don't even
bother trying any more. If something is not in the latest Platform
version, I just find ways not to use it
* The Macport's portfile is outdated for the haskell platform, and it is
still at 2009.2.0.2
* Library standards: namespaces, general quality. The Haskell Platform is
a good start, but Haskell needs to go even deeper.
* The Haskell Platform is also a _huge_ win and has made my Haskell
* No PPC support on new GHC/Platform releases.
* The haskell platform should span over a greater set of libraries, which
should be guaranteed to work on all platforms on each release.
* Great language, needs focus on libraries (haskell platform is a good
start)
* Also, the multi-platform issue is a big stopper for me. I'm developing
on Linux and Windows, maybe on Mac. Mac supports is simply broken,
Haskell platform does not install. On Windows, many packages relying on
C libraries simply do not work. On Linux, Ubuntu does not have official
up to date packages
programming life much simpler.
More information about the Haskell-platform
mailing list