[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Maintainers, porters, and burden of porting

Le Wed, Aug 31, 2011 at 11:35:59AM +0200, Andreas Barth a écrit :
> * Lucas Nussbaum (lucas@lucas-nussbaum.net) [110831 10:56]:
> > Also, in the case of architectures targetted at embedded systems (I'm
> > thinking about mips and mipsel), what is important is that Debian
> > infrastructure supports the development of those architectures, but I
> > don't think that there's much to gain by being officially supported if
> > it's only used in production through derivatives that can provide the
> > official support.
> You are aware that there are mipsel netbooks? And arm tablets?

In some previous discussion I was also pointed at 64-core mips workstations.
How many of these machines are running Debian ?

I do not think that we can consider ports equally.  The arm platform and the
armel port have some clear success.  According to popcon, the user community of
Debian on armel is constantly growing, and is aproximately 1 % of our ‘PC’
(i386 and amd64) user community.  Also, other distributions, for instance
Ubuntu, increase their committment for this platform.  In comparison, the
usrerbase of the mipsel port stagnates to 0.05 % of our PC userbase.  If there
were many Debian users of mipsel netbooks and workstations, why would they not
use Popcon, as the Debian users of armel computers do ?

Having Debian running on rare architectures is a great acheivement.  However,
overestimating their user base results in turning maintainers and porters
against each other.  While porting a program on a rare architecture will raise
its quality, there is usually no shortage of bugs that directly affect users on
more popular architectures, and solving them also raises the program's quality.

Importantly, our current practice is not raising Debian's quality.  In
contrary, it makes us distribute a large number of packages that do not work at
all.  It is usually not noticed because nobody uses them anyway, until the
maintainer activates some build-time regression tests.  Many specialised
packages have insufficient testing on the architectures that are not reported
to be used upstream nor in their users community.  I think we would benefit of
a system that acknowledges this, and should have more flexibility for taking it
into account when deciding where we spend our time.

Have a nice day,

Charles Plessy
Debian Med packaging team,
Tsurumi, Kanagawa, Japan

Reply to: