Re: Debian rolling: tentative summary
Pierre Habouzit <email@example.com> writes:
> On Mon, May 02, 2011 at 01:31:31PM +0200, Lucas Nussbaum wrote:
>> Since I already sent too many mails in the 'rolling' discussion, I
>> decided to send one more. Here is an attempt at a summary of what was
>> said so far. It might not be complete, it's probably a bit biased, but I
>> hope that it's still better than nothing. When replying, please try to
>> focus on specific points, and change the subject accordingly.
> That's a decent summary of what was said I think.
> Though I feel that to make the discussion more solid, the following is
> - What are the problems you try to address with rolling? And no "the
> users want it" isn't an answer, I'd reply "why do they want it" if
> that's the answer I get.
I think "why do they want it" is the most important question.
As mentioned many users feel that the software in stable is too old. On
the other hand many users praise Debian for their long release cycle and
stable/old-stable support. Those two groups are clearly expressing
contradictory wishes. And maybe trying to find a solution that fits both
isn't the right thing.
To those users that want newer software my next question would be "What
software?". My feeling there is that it is only some software, allways
the same software and used for the same use case: KDE / Gnome /
Multimedia stuff for the Desktop.
Looking at the other group, those that cherish the slow release cycles
and excelent stability and security support, shows a a large bias
towards servers that are either completly headless or where people don't
care about the latest sparkly desktop hype. They need to run and keep
running without having to upgrade them every 6 month.
If my impression is right then maybe there is something to say for
having a desktop and a server flavour like other distributions. It would
be wastefull to have a rolling release with all sources included if the
users only need a subset. The desktop users only want the new sparkly
KDE / Gnome / Multimedia stuff. They do not care about the latest
coreutils or latest postgresql.
So my idea would be to split pakages into 3 groups: core, desktop and
server. We could then have "full" releases that update all sources
(core+desktop+server) and "sparkly" releases that only update desktop
[OK, so only 2 groups: long-term and short-term]. The desktop packages
for sparkly releases would be build against the core packages from the
last full release. They could be done as rolling releases or not. That
option remains open. The point would be to greatly reduce the number of
package updates involved in a sparkly release, to reduce it to those
packages that matter, and therefore reduce the work needed to pull it
Note: yes, this isn't exactly like other distributions having a desktop
and server flavour. You would still have all the packages available in
every release. The up-to-date-ness would differ, not the amount of