Where are we going? (Not a rant)

[Michael Scherer]

Le samedi 08 décembre 2012 à 19:31 +0100, Ralf Corsepius a écrit :
> On 12/08/2012 05:31 PM, Michael Scherer wrote:
> > Le samedi 08 décembre 2012 à 05:12 -0800, Dan Mashal a écrit :
> 
> > In fact, I never heard anyone complaining about "kde is dying" while the
> > numbers are much more worrisome :
> >   http://www.ohloh.net/p/kde/contributors/summary
> 
> 
> > Maybe that's caused by some cleaning of their svn, or maybe that's just
> > nokia's problem ( and the huge layoff that followed ).
> >
> > I would also add that if the switch to gnome 3 made enough people leave
> > the project,
> Which project are you referring to?

I refer to GNOME, as "the project" from  
"For example, the same thing happened with Gnome 3 upstream where a lot
of developers left the project due to a lack of a real vision or
direction."

It was obvious in my head, but in my head only :/

> >> "Let's make it look like Ubuntu because Ubuntu is popular" is another philosophy that
> >> I have seen and I have a problem with.
> >>
> >> In addition, LTS won't solve this problem. It goes against everything Fedora stands for.
> >
> > So "first on feature" is a problem, but that's what Fedora stand for.
> > Yet, LTS would be bad because it goes against what Fedora stand for.
> 
> IMO, no. LTS can also mean to improve parallel installations (ship 
> forward-compat packages).
> 
> An alternative to LTS could be to improve upgrades (Make them 
> easier/seamless) and encourage backward-compat packages.
> For instance, I never understood why "last RHEL"-compatibility had never 
> been an objective for Fedora.
> Though this certainly will not always be possible, achieving this is 
> likely possible for a larger number of packages.

There is a chicken and egg problem. To be sure we are still compatible,
we need people to test that. But no one will test if we are not
compatible. 

And I do not deny this would surely be helpful but it can quickly become
complex as hell. 

If I take for example openshift origin, running on RHEL 6, being able to
run it unmodified requires to keep older version of rails ( as the
upgrade to rails 3.2 broke some code ), older ruby ( as I am not sure
that all gems support ruby 1.9, but that's just a speculation and i may
be wrong ), and surely a few gems. This also requires keeping some
compatibility layer for sysctl knobs, cgroups, etc.

And then we are back on the software collection discussion, among
others. And this would also conflict with the idea of not diverging too
much from upstream, if upstream do not care about compatibility. 

This would also mean that lots of thing would break every 2/3 years for
each new RHEL version, and that we let Red hat decide of the
compatibility of Fedora, which is undesirable IMHO.

Another problematic area would be the kernel. Do we want to have RHEL
compatibility on this ? Or Xorg ? 

-- 
Michael Scherer