Ac stable on our doorstep?

[Marcin Król]

> Keep in mind that Ac already has obsolete versions of gcc, libstdc++, glibc
> and X11. Changing each of them would mean big revolution.

Thats true. But its still possible to maintain distro this way.

> So it's natural to maintain just necessary updates. And maybe some
> individual packages which are easy to update without rebuilding half of
> a system.

With "always in developement" way we of course will need milestones for
releasing big changes like those mentioned above. Others do that, why
can't we? In my opinion "aid" way should cause releasing
snapshots/milestones more than once per 2-3 years.

Ac will be released this year. One year from now it will be possibly
outdated. And I'm pretty sure that Th will not be released before 2008
if we will keep current way of developement.

So in my opinion we have one year for changes before we will get back to
state where stable release will be outdated and devel release will be
far from stable.

> If there is a need for more stable system with newest packages, XGL and
> so, it's time to stabilize Th and stop experiments such as gcc snapshots,
> development versions of widely used libs/programs (glib, gtk+, python...).
> If experiments are wanted - remember nest? It could be resurrected.

Thats a good idea for the beginning. Moving all experiments to nest will
give us almost exactly what I'm thinking about:

- Ra, Ac, Th would become milestones ;)
- by taking out experiments from HEAD we will be able to release
milestones more often
- more often milestones will allow upgrades from previous ones (fastest
way of Ra->Ac migration was reinstallation)
- life time of milestone release will be shorten, so small number of
developers maintaining its updates will no longer be a problem
- we won't be stuck with old software for 2-3 years (like in Ra) before
next version becomes usable

M.