Request For Candidates: Application Review Board

Allison Randal allison at ubuntu.com
Tue Aug 17 02:11:20 BST 2010


To take a step back to the broader vision (and answer several parts of
this thread at the same time):

This all started at UDS a couple of years ago, in a meeting about how to
attract more application developers to Ubuntu (and Linux in general). I
offered Scratch (scratch.mit.edu) as an example of how easy it can be to
develop and distribute applications, and of the kind of community energy
you can build up when you lower the barriers to entry. Three days later,
Quickly (https://wiki.ubuntu.com/Quickly) was born.[1] Quickly has done
a fantastic job of solving the "develop" side of the equation, making
sane default choices, common boilerplate starting points, and reusable
widgets for developers. The "PostReleaseApps" process is the next
logical step, looking at the "distribute" side of the equation.

Since that first meeting, the iPhone/iPad App Store and Android
Marketplace have clearly shown that there's demand for revolutionary new
ways of distributing lightweight apps. User expectations are shifting,
and so are developer expectations. Now those same iPhone apps are
running on the iPad, and it's only a matter of time before the Apple
desktop is dominated by the same (or similar) App Store. It remains to
be seen if Microsoft will catch the wave, but I've had conversations
with others who see the writing on the wall for Windows too, and want a
slice of the pie.

What we have now in the PostReleaseApps process is a very conservative
toe-test of the waters. Several comments in this thread are around the
general theme of "How is this any different/better than REVU?" Well, it
really isn't yet. There are important security and quality reasons why
our current packaging process is what it is. It's a solid, reliable
process and we won't diverge far from that in the first round. But
there's great potential in the future. For example, Android and Scratch
can be so completely open to new app distribution because the code runs
in a tightly controlled sandbox, with guarantees that the worst a bad
app can do is crash itself (think "PyPy sandbox" but better).

There's been great feedback in this thread on things like the
distinction between the current distribution process and this new
process (not enough distinction yet), on the review process (still too
heavyweight), demands on our limited developer resources (still too
high), the developer process for building the packages (still too
heavyweight), handling dependencies and porting to later releases (still
needs architect/design/implementation work).

The design and development of this idea will happen on this list and on
#ubuntu-devel. (There's little risk of overwhelming the general traffic
here, and a great risk of missing good feedback if we created new
channels.) Thanks for your help shaping it.

Allison

[1] I wasn't directly involved in either Quickly or PostReleaseApps. I
just started last week as Technical Architect of Ubuntu. But, from
another perspective, I also feel responsible for sparking the idea in
the first place, both proud of how far it's grown already, and conscious
of how much potential still lies ahead.



More information about the ubuntu-devel mailing list