Re: GSOC 2008 advice

["Luis Villa" <luis tieguy org>]

On Tue, Feb 26, 2008 at 1:10 PM, Benjamin Gramlich
<benjamin gramlich gmail com> wrote:
> Greetings all,
>
>  I am interested in applying to work on a project for Gnome during the
>  summer of code 2008, and I have a few ideas.
>
>  Idea #1) Re-implement the panel-applet library/interface to depend on
>  DBUS.
>
>  Idea #2) Migrate the panel to GIO/GVFS and DBUS.
>
>  Idea #3) Develop a tutorial for GIO/GVFS.
>
>  Idea #4) Create more compositing effects for metacity and develop a gui
>  configuration tool for the effects.

Great to hear you're excited, and great that you've come to d-d-l for
advice! Some advice from someone who hasn't been involved in SOC for
several years now, so take with a grain of salt :)

* "widgets": Vista, OSX, and KDE4 all have widgets/gadgets/Kthingies
that are pretty, very easy to use, very easy to develop (since they
are web-based), and which display more information when needed while
staying hidden when not needed (both unlike our panel applets.) Some
work has already been done on doing this with gtk-webkit[1]- perhaps
that could be built on? (It seems to me that from a user perspective
this approach is really superior to applets and what we should be
focusing on long-term instead of reworking applets, but YMMV.)

* panel: there are two promising new panel alternatives: bigboard and
the combination of AWN and gimmie. Bigboard hasn't had public releases
in a while[2], so I'm not sure where it stands, but there is some very
interesting gimmie-AWN work going on[3]. Helping one of those projects
get in shape for use by 'normal' users would be interesting and
productive work.

* menu/run-dialog: Windows and KDE4 both have search and MRU deeply
integrated into how users find and run applications, and it makes them
simultaneously easier and more powerful to use. These features have
been prototyped but not deployed into mainline GNOME, via deskbar[4]
and Novell's 'slab'/gnome-main-menu. Diving into those projects,
figuring out what prevents them from being mainlined, and working on
those rough edges may be worthwhile. (May not be worthwhile; the big
problem for deskbar may be python rather than anything that can be
solved on the gnome side, and g-m-m seems to lack a webpage, so I have
no idea what the status of that project is.)

* mozilla 'prism' integration: gratis-free webapps are a big part of
the future of the desktop; making sure they integrate well into GNOME
should be an important task. So investigating Mozilla's prism[5] or
similar technologies, and doing a good job at tying them to GNOME
(e.g., currently prism places a launcher on the GNOME desktop but not
in the menus) might be interesting and worthwhile, and educate you
about two codebases instead of just one. This may also tie into the
bigboard work mentioned above.

* people browser: this may fall into the category of chasing
windmills, but there is some hacking going on in soylent again[6] and
I think it would be cool ;)

* testing framework: there has been recent discussion here on d-d-l
about testing frameworks[7]. They aren't glitzy or glorious, but they
are critical. Working with the QA, buildsquad, and a11y folks to
evaluate the three competing technologies and ensuring that one of
them gets run automatically every single day would make you a very
serious hero with an impact all over the project.

So... those are my suggestions; like I said, take them for what they
are worth, but I think they'd be a worthy starting place for anyone.

Luis

[1] http://www.atoker.com/blog/2007/10/29/gtk-matters/ - see screenshot mid-page
[2] last release of new tarball built for fedora was in October:
http://koji.fedoraproject.org/koji/packageinfo?packageID=4385
[3] http://groups.google.com/group/gimmie/browse_thread/thread/6bbc370656777f47
[4] http://raphael.slinckx.net/deskbar/
[5] http://labs.mozilla.com/2007/10/prism/
[6] http://svn.gnome.org/viewvc/soylent/trunk/?view=log
[7] http://mail.gnome.org/archives/desktop-devel-list/2008-February/msg00103.html