Freeze! Some lateral thinking here. ( was: Debian decides to adopt time-based release freezes )
Hello,
should we possibly think of alternatives to the current way to present releases? I was
thinking about the possibility to adopt some principles from version management that we
got accustomed to in programming ...
My hunch is that everyone has a few tools that he/she wants to work on in daily routine.
Those need to work no matter what - like the window manager and what is underneath. The
user will be reluctant to change anything there unless there is a vulnerability of some
sort, for instance. But that should be a smallish update only, no major release, and a
user may have a default setting to allow smallish important updates, but object larger or
non-important updates.
Then there may be some other kind of software that shall be always at their latest
version. Say, the boinc-client, which does not matter much if it temporarily fails or one
goes back to the previous version.
What today is a release, would then be a tagged set of packages, possibly implemented as a
set of symbolic links to some large package archive. A new user would get that as a seed
of packages. From then onwards, the apt tools would suggest only packages that are
smallish updates and important to substitute the currently installed version. Upon manual
initiative, the user could select newer versions for selected packages when feeling ready
to evaluate something new and risk a temporal failure of some sort.
Does that make any sense to you? The maintenance of a release would then mean to continue
providing smallish important updates.
We would still need a release team, obviously. But we would possibly concentrate more on
the core functionality for defining a release, like the libc etc. When talking about a
particular installation, then we would then describe it like "lenny + KDE 4". This would
develop the "we release when it is ready" more towards "the users accept releases when
they are ready". And, I can imagine that many more users, who are often power users of
some sort, would use that opportunity to be very close to the package maintainers and to
upstream while using the packages from testing or unstable, since they just do not risk
that much while upgrading selected packages to the cutting edge. This would help our
communication a lot.
Best,
Steffen
Reply to: