[Mageia-dev] Release cycles proposals, and discussion

[Michael Scherer]

Le mercredi 15 juin 2011 à 02:33 +0900, Jehan Pagès a écrit :

> >> And as I said in another mail, if people want to follow arch linux and
> >> do a better job, maybe they should start to explain what are the weak
> >> points of the distribution and then do proposal on stuff that can be
> >> done better instead of asking to copy cat hoping this would be better.
> >
> > I don't want a full rolling release, because of the listed disadvantages.
> > So, if you ask me what is "wrong" with Arch, I would say:
> > * due to the rolling release, it's nearly vanilla. This doesn't match
> > requirements of Mageia
> > * no innovations (because of vanilla)
> > * a rolling core system has a negative correlation with it's stability
> > * heavy work load
> > * ...
> >
> > So, I don't ask for a copy of Arch nor any other distribution. I asked
> > (although it wasn't my idea) for something new. An compromise: a light
> > rolling release.
> >
> > Further lack of clarity?
> 
> So basically what people call a "light" rolling release in this thread
> is a rolling release where packages are tested and integrated? And
> what you call a (non-light) rolling release is a development rolling
> release (cooker, cauldron…) where packages are just dropped without
> prior security checked as fast as they are made available by their
> respective authors?

Then light is what arch does. They do tests before migrating. Now, this
can be done faster because the integration is minimal, per philosophy
( ie, you take care of configuration and fixing everything ). Heck, we
could even fully automate that with mdvsys upgrade.

> If so, I would say, yeah obviously "light" (I find this naming quite
> paradoxical then) is the kind of rolling I would like. And that's not
> that new, that's the kind of rolling release in Gentoo (which I found
> much more stable than my years of experience in Mandriva, and also
> more peaceful as I don't have to fear the big update every 6 months
> which will definitely break a lot of small stuffs everywhere at once).

Having lost access to my server for a whole weekend due to a library
upgrade on Gentoo ( some stuff linked to libncurses or libreadline, so
no more shell ), I beg to disagree.

Also, gentoo seems to be updating more slowly nowadays :/ hence the
stability

> Also yes, I guess this could be simulated using the current backport
> system becoming a supported repo (with package getting appropriately
> tested and the right integration into the distribution done). I don't
> say this is the ideal system, but that can be a first step in the
> evolution.

The problem is that the more testing we add, the more ressources it
requires, and the more time it requires. And soon, people will complain
that "packages are too old". 

But if that what people want, we can have the same QA for backports and
for updates. But then I hope there will be many testers.
-- 
Michael Scherer