Re: [Caml-list] Some comments on recent discussions

[Gerd Stolpmann <info@gerd-stolpmann.de>]

Hi Martin,

GODI is currently broken on Windows, and I would need to invest again a
few days to get at least the basics running again. This is a big problem
for me, because I've personally no direct advantage from this, and there
is also the question how you can keep such a very different port running
when you don't use it (i.e. errors will start accumulating again). At
least I'd need support from a skilled porter with direct interest in it.

Also, the usefulness is not totally clear to me. Actually, only those
GODI libraries will run painlessly that are either ocaml-only, and that
have been specially ported. So for any bigger real-world app you would
clearly run into big problems anyway, because it is almost sure that
there is a non-ported library in the dependency chain. It might be good
enough for education or programs not needing deep OS support, but it is
clearly limited.

Gerd

Am Montag, den 12.12.2011, 21:54 -0800 schrieb Martin DeMello:
> Better Windows support would be very nice too. A friend recently had a
> python app that he wanted to port to native code, and I offered to do
> it for him in OCaml. The linux version was quick and easy to develop,
> and we both believed that he could just install OCaml and the required
> libraries on windows and have easy cross-platform support with little
> effort. He spent two days trying to get GODI and Batteries working,
> and finally gave up; it was eashobier for me to just rewrite the bits
> I was using directly in my code and not have dependencies. For a
> larger project, it really wouldn't have been feasible not to be able
> to rely on GODI, and (since my main area of interest is end-user
> native apps) this has made me wary of using OCaml for anything other
> than apps for my own use, or apps that I am very sure will only be
> used under linux.
> 
> martin
> 
> On Tue, Dec 6, 2011 at 7:24 AM, Jonathan Protzenko
> <jonathan.protzenko@gmail.com> wrote:
> > Dear OCaml hackers,
> >
> > I'm very uneasy about the current opinions that are voiced on the caml-list.
> > I have good reasons to think I'm not the only one in that situation, so
> > please allow me to raise a few concerns about some recent discussions.
> >
> > There's several subtopics in the "OCaml maintenance status / community fork"
> > that I'd like to discuss.
> >
> > = Improving the community =
> >
> > I think the main point of the discussion is to improve "the community". If
> > we really want to improve OCaml as a whole, then I think we can put our
> > efforts on better areas than patching the compiler.
> >
> > == Package management system ==
> >
> > The thing that's most needed is, imho, a package manager that works.
> > Oasis-db looked very promising as far as I could tell, but Sylvain doesn't
> > have as much time as he used to do. Instead of hacking on our pet projects
> > (which is, I admit, very rewarding), maybe someone could step up and make
> > Oasis-db happen. We don't have a single, unified answer to "what should I
> > install to easily hack with OCaml?". What made Python, Perl, Haskell
> > successful is the package management systems. How much longer are we going
> > to shy away from this issue? Sure, it's much more fun to hack on the
> > compiler. Not as useful.
> >
> > == Leaving our own corner of the web ==
> >
> > The OCaml community likes to stay in its own corner of the web, in
> > isolation. We live on obscure web sites: who knows about ocamlforge outside
> > the OCaml community? Who knows about the caml hump? We could host our
> > projects on Sourceforge or on GitHub. We could get recognition in the
> > open-source world through our projects, we could be more social, we could
> > boost the language stats on ohloh, we could attract more contributors (being
> > a fervent user of GitHub, I must say I've attracted a significant amount of
> > contributors that way ; being on an obscure forge, I'm certain it would've
> > never happened). We stay away from that. Why? Because GitHub is not
> > open-source. The whole point of git is that everyone, everywhere has a
> > backup copy and that we don't care if GitHub falls down. Nevermind.
> >
> > GitHub has a fantastic integration between the bug tracker, the commit
> > messages (git commit -m "Fix #486" closes bug 486 on the bug tracker), the
> > source repositories. You can discuss patches in-place. You can interact in a
> > very easy manner. You can do peer-review in a snap. GitHub is the only place
> > that leverages the social nature of open-source collaboration! That's
> > precisely what we're trying to achieve here. Are we going to throw all that
> > goodness away? What kind of signal are we sending, when we're considering
> > hosting our own instance of gitorious? A very simple one: "we like to stay
> > in our own corner — don't come".
> >
> > There are valid points: GitHub doesn't have a maliing-list list system. Is
> > that really an excuse? How hard would it be to setup a website that lists
> > OCaml projects, provides them with a mailing-list, and points to their
> > GitHub page?
> >
> > = What is this about ? =
> >
> > If it's about improving the general situation with OCaml and its community
> > (the title of this thread contains the word "community"), then I believe
> > hacking on the compiler is not the most effective way to achieve that goal.
> > We're hackers. We like to hack on things. And we often fail to ask
> > ourselves: is it really worth implementing? Submitting patches is easy.
> > Submitting quality patches that do solve a real problem is harder. The ARM
> > backend does need a cleanup, and the patch does solve a stringent issue.
> > That may not be the case for all patches.
> >
> > There is indeed a problem w.r.t external contributions. I agree that the
> > INRIA team could make it clearer what its stance on external contributions
> > is. Maybe one solution would be to have a INRIA-endorsed ocaml-next on
> > github that everyone can fork, where we would merge really outstanding
> > features, before submitting them to INRIA, as you described. I don't think
> > it is such a good idea creating a real fork. Maybe some sort of integration
> > platform on GitHub would be the right solution to the "patch review"
> > problem.
> >
> > I'm not even sure what kind of patches you wish to see integrated. Can you
> > clarify that?
> >
> > = Conclusion =
> >
> > This is indeed a long rant, but I'd like to see us being more practical and
> > down-to-earth. I love OCaml. I think we can do better for the language.
> >
> > Kind regards,
> >
> > jonathan
> >
> 
>