On Mon, 2008-10-13 at 14:29 +0200, Patrice Dumas wrote:
On Mon, Oct 13, 2008 at 01:18:01PM +0100, David Woodhouse wrote:
>
> I understand that there is a market for a Fedora-based distribution
> which doesn't receive megabytes of updates each week, and which is
> supported for longer than a year.
What have been proposed so far doesn't solve the 'which doesn't receive
megabytes of updates each week' part, only the 'supported for longer
than a year' part.
> What I _don't_ understand is why these requirements are not met by
> CentOS. Isn't that _precisely_ the 'market' that RHEL and CentOS exist
> to serve?
Because it is not the same as you stress yourself. Centos is clearly not
the same than 'Fedora-based distribution which is supported for longer
than a year'. It allows to use innovative technologies while not being
forced to update each year.
This is the part I have difficulty understanding. You want to use new
and innovative technologies, but you don't want to update your nice
stable system?
Do you want a perpetual motion machine with that too?
You're right -- you are offered updates, or you are offered stability,
and there isn't much of a middle ground. But that's just reality.
> As I see it, there is a continuum of sorts -- from the daily
churn of
> rawhide, through the less anarchic but still considerable churn of the
> latest Fedora release (currently F9), to the more conservative set of
> updates for the previous release (F8), and then a bit of a jump to the
> long-term stagnation¹ of RHEL/CentOS. You can pick whichever one you
> like, according to your needs.
The last jump is not realistic in all cases. What should F6 user jump
to? Centos 5? Centos 6? And F8 users?
F6 is basically RHEL4/CentOS 4 for the most part, isn't it?
And F8 would probably be closest to RHEL5/CentOS 5.
The proposal is not to create a new distribution, but simply have
EOLed branches acl removed and leave the possibility to build and push
the results, with a limitation of the changes to grave bug fixes and
security issues.
So no new innovation then? And you'd want to do this for _every_
six-monthly release of Fedora? Surely that's a whole boatload of effort
you don't need? Why not just do it for every other release? Or, perhaps
more usefully, every third release -- a new one about every 18 months?
I see no fundamental reason why we should _forbid_ people from doing
security updates for packages in EOL distributions, although I'm very
wary of the expectations that it would create. If we ship
official-looking updates for _some_ security bugs, naïve users will
quite reasonably expect that they'll be receiving updates for _all_
serious security bugs, and our "You are unsupported; you need to upgrade
before you get hacked" message will be compromised.
I also think that with a niche market that small, between Fedora and
CentOS, you are unlikely to get enough volunteers to keep it viable --
isn't that what happened with Fedora Legacy last time?
(That being said, it is also possible that 'which doesn't
receive
megabytes of updates each week' issue may deserve to be looked at, but
it is a different issue).
That and/or ensuring that the packages you miss in CentOS are actually
in EPEL.
--
David Woodhouse Open Source Technology Centre
David.Woodhouse(a)intel.com Intel Corporation