[haiku-development] Re: Discussing GSoC 2015 Ideas

  • From: Adrien Destugues <pulkomandy@xxxxxxxxx>
  • To: haiku-development@xxxxxxxxxxxxx
  • Date: Mon, 26 Jan 2015 11:46:03 +0100

On Sun, Jan 25, 2015 at 05:09:42PM -0500, Augustin Cavalier wrote:
> On Sun, Jan 25, 2015 at 4:44 PM, Andrew Hudson <dmarc-noreply@xxxxxxxxxxxxx>
> wrote:
> > I think Haiku 64-bit apps are absolutely perfect for GSOC and there are
> > several ways to approach them. Each applicant accepted could start with a
> > porting basic app,
> > there are a number that Scott has listed in Haiku Ports. The student could
> > start with one app, get it ported to Haiku B1, build a recipe, get it on
> > HaikuDepot, do
> > the same for Haiku 64-bit. And then start their project as a whole. Some
> > students might just want to work on ports all summer. If they ported one or
> > two apps a week,
> > that would be amazing, and also excellent GSOC experience. Can you imagine
> > the benefit to the Haiku community as a whole? That's a huge benefit to
> > the Haiku community as a whole.
> >
> As I said before, this is not hard at all, and does not consume much time.
> As a "community bonding" task yes, as a real project no.

In general, GSoC is about getting people to contribute to Haiku. In some
special cases we put them to work on 3rd-party applications or
libraries, for example we did this for SDL2. In these cases, we had
support from the SDL2 developers who helped mentoring the student.

GSoC isn't really about getting the tasks complete, this would be done
much faster if the mentors spent their time coding. Instead the goal of
GSoC is to get some of the students to become and stay Haiku
contributors. We can't do this by asking them to work on external
projects we have no idea about.

Porting apps to Haiku using haikuporter is an extremely boring task for
average developers. Yes, a student could get through a lot of them
during a summer of work, but there is no benefit to that for the
student. He does only very boring work, he does not get in touch with
the respective apps development teams (because he would at best submit
them just one patch and then go away), and at the end of the summer, he
will move to something else. Nothing is won here.

> >  Regarding Blender - I don't think it only does what you think it only
> > does. Yes, it has a hardware acceleration part to it, but with multiple,
> > modern CPU's
> > it could render simple scenes in real time, complex scenes in near real
> > time. What Blender for Haiku could use is a Haiku-specific viewing layer
> > that would
> > render directly to the display buffer in a video card.
> >
> BGLView can be used in a BDirectWindow, yes. However, no, simple scenes
> would be stuck probably at 30fps, with complex scenes dropping to <1fps. My
> brother uses Blender extensively on Windows with a mid-line graphics card,
> and sometimes it'll drop below 30fps. In animation (~50% of Blender users),
> anything less than 30fps is no good as you can't accurately judge what the
> rendered version will look like.

There is this, and also the fact that Blender was already ported to BeOS
some years ago. Which turns this to yet another boring recipe-writing

> And who would mentor this? Who is around that has a lot of knowledge of the
> codebase?
> >  Regarding IPv6. IPv6 is so much more than just a kernel module. I would
> > recommend that you take a college level networking class before weighing in
> > on this topic.
> >
> I know what IPv6 is, thank you very much.
> > Here is a link to some apps that could be ported and/or tested as part of
> > Haiku's IPv6 adoption and compatibility:
> > http://en.wikipedia.org/wiki/Comparison_of_IPv6_application_support
> > This could easily be a graduate level project. Again, this would be
> > incredible on someone's resume and would a huge benefit to the Haiku
> > community.
> >
> Not really. As mentioned in Adrien's progress reports, IPv6 is really just
> missing proper DNS resolution, which he's working on. Other than that, it's
> already complete.

This is wrong, however. DNS for IPv6 already works (all too well
actually). We even get all apps to try connecting to IPv6 hosts even
when there is no network interface configured with IPv6. What I've been
working on is a fix to that, so we can stay in IPv4 world safely.

I don't currently plan to do more work on the IPv6 side, as, while I
know what "it is", I don't have much knowledge on how it is implemented,
the various protocols involved, etc. I could dig into that but some
other members of the development team will probably know more about

So, this is not about shouting random ideas of "nice to have" apps.
First because GSoC should rather focus on writing code for Haiku itself,
and second because what we need is mentors to go with the ideas. Please
don't offer ideas if you don't think you can mentor them yourself, or
know someone who can and wants to do it.


Other related posts: