When the release drags on so long, new upstream releases are made and developers want to package these so they can use them. e.g. gnome 2.6, gnome 2.8...
There are two options:
- Break sid, delay the release
- Use experimental.
Experimental is a pain.
- It often has experimental packages in it which are broken
Users might want the latest gnome, but not at the risk of breaking other packages. Having separate, autobuilt experimental repositories (experimental-gnome, experimental-utopia, experimental-gcc...) would fix these problems.
This is technically possible as experimental IS autobuilt.
Note that gnome does work well in experimental -- proving that the concept is good. The problem is that while the gnome team has done a lot of work to make this work, the average developers don't have the time to do this for their own smaller packages/groups of packages.
What is the problem with only pulling certain packages from experimental using AptPinning? It is kind of a pain since /etc/apt/preferences doesn't support wildcards in package names, but at least you don't get the entire amount of possibly broken experimental packages.
experimental packages are built against unstable anyway, so there shouldn't be any problematic dependencies -- ZugSchlus 20050601
Back to ReleaseProposals