[Mageia-dev] How will be the realese cycle?

[Michael Scherer]

Le lundi 04 octobre 2010 à 20:35 -0500, Fernando Parra a écrit :


> One of the regular questions I received when install a linux distro is: "But I really need to upgrade all my OS, if I only want the new version of my Office Suite?"
> 
> Let me remind you, The privative OS don't have this model, 
> in them the users can install a new version of their favourite 
> software as they want (or pay it). This is true at least as a 
> Mayor upgrade at the OS, and in some cases the compatibility are maintained.

Well, their model is too different, I have made a full 40 minutes
conferences on the subject, except it is in french ( and the video is
still not there ). 

But basically, that's because Microsoft or Apple work in a segmented
fashion. You have clear delimitations on what is the os, and what is
not. Os is controled by a single entitie, and there is only a few APIs
that you can use, and lots of test to ensure that the API is not broken
( even if that mean keeping old code around, see
http://www.joelonsoftware.com/articles/APIWar.html ).
This segmentation also mean there is few lines of code put in common, be
it for technical reasons, or legal ones. And the lack of code sharing
mean complexity of the source code, which usually also mean more bugs,
and more time to develop ( thus, less quality, if you invest the same
ressources ).

Never wondered why windows takes so much memory ? Now, you have the
start of a answer.

Now, on free software side, there is no artificial barrier between os
and applications, no committee to declare "this is the official API for
X years". There is just a bunch of people who constantly give source
code, with the hope that everything will work fine together. And that's
what we do as distributors, we pick components everywhere and make sure
they are presented in a way that everybody can use. We even make more,
we make sure they integrate cleanly with the others ( menu, code and
library sharing, packages system, integrated installation ), which is
something that usually is not done in proprietary os.

People could do like windows. Just take a distribution, declare that you
will never upgrade it for 4/5 years. Be sure to have a team ready to
backport security fixes, and you will have the stable plateform you
want.

But then, people will want the latest kernel, to get support for their
hardware. The latest xorg, for the same reason. But then, after
upgrading xorg, you will need to update udev, and hal. After that you
will see that gnome or kde no longer have a working automount so they
need to be patched, or updated. But updating gnome may requires to
update others components ( like gstreamer ) and so on.

Even with this problem, this kind of distribution already exist. That's
RHEL, or Centos. ( Or even Opensolaris, when it was still existing, or
maybe others ).

But :
1) RHEL is not gratis, far from it. Centos is, but I do not think Centos
could exist without RHEL. That's because the whole "let's validate the
API" business is costly for the distribution. Far more costly that what
we can afford as a community distribution, and far from being sexy for
coders. For the coder that use a old API, this usually mean the API
offer less features. For the coder who wrote a library and thus offer a
API, this usually mean that no one test your new and improved code,
which can be simpler, easier to debug, etc. So you cannot count on
"free" coders, ie you have to pay someone to do the job. You also have
to pay someone to do the formal and rigorous QA.

2) newer distributions are usually offering more features. And people
want features. The start of this discussion is "how can I install the
new office suite", which basically mean "i would like to have new
features of some software". While everybody will tell "but I only need
this feature", nobody will give the same feature.

But I think people can try to do it. Just install Centos to your
friends, and make a rpm repository of update to firefox, to openoffice
and others, and see how it goes. Or maybe use a BSD, like PcBSD
( http://www.pcbsd.org/ ).

Maybe you will have success. And maybe not.

Personally, and as arrogant and elitist it sounds, I value more the
innovation than binary endless compatibility, and so I think free
software ressources are best spent into new features than into keeping
unmaintained binary softwares working ( because yes, a software that no
one recompile since X years is a unmaintained piece of binary to me ).

So, to me, the current situation is fine. I know it is not for
everybody, but as I said, for people who want a different system, there
is a solution. 
But I do not know why people do not use it. 
-- 
Michael Scherer