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

Re: Status of Potato

On Sat, Nov 06, 1999 at 10:12:27AM +1100, Chris Leishman wrote:
> ATM we work hard to build a good, solid distribution - which we then release
> to the public.  Once it is released we all have a party, a few drinks, then we
> leave it and move on (starting a new "unstable").  Because we a part time,
> volunteer developers, we can't be continually supporting three different
> development trees.  Instead, we concentrate our efforts on one - and we make
> it the best in its class when it is released.  To try and do more would put
> me, and many other developers, out of the project.

I understand your arguments but, as a user, also feel Sami's pain. :-)

I use only the stable release on several important servers.  When a
server is to be upgraded, I build up a replacement on new hardware, hack
and bash away on it, test, re-install from scratch, re-test and then
replace the old server outright.  During the hack-and-bash phase, the
new server often includes needed packages from unstable, in anticipation
of a release in time for final testing.  Sometimes I can't wait for a
release and build local versions from the upstream source and selected
parts of the maintainer's diff.

Running hybrid stable/unstable systems is painful.  Trying to install
a single package from unstable brings a cascade of dependencies on
other unstable packages.  Some of these dependencies seem excessive:
a shell script in an unstable package calls bc and therefore the package
depends an unstable release of bc, which depends on an unstable glibc ...
Obviously unstable packages are generally built on unstable, but it would
really be helpful if the dependencies were from stable, where the stable
packages were, of course, suitable.

On a related note, it would be nice if the package tools knew about
packages from both stable and unstable at the same time, allowing a user
to select the unstable version of a package and see right away what else
needed upgrading to support it.

For me, hybrid systems fill the gap between stable and unstable while I
am developing new servers.  I suspect I am not the only one to operate
this way.  Making hybrids more comfortable could benefit many people.

Thanks for listening,

Mark van Walraven
Wave Internet Services

Reply to: