It's unsurprising - and commendable - that community members want to eat
their own dog food when promoting their favorite Linux distribution —
but sometimes the dog food expires far too quickly. This was brought to the
fore, again, by a question posed to the Fedora Board about Fedora-it.org, which is running on a hosted
server running Ubuntu.
On January 9, Gianluca Sforna queried
the Fedora Board about the Italian community site for Fedora, which is
hosted by Italian Linux Distro Network
(ILDN). Sforna wanted to know whether the project has, or wants to have, a
position on the topic of hosting a Fedora site on a non-Fedora server
— not an idle question, since the site is in a development stage and will
soon be seeking approval from the Fedora Board to use the Fedora
Former Fedora Board member Matt Domsch quickly, and many would say reasonably, replied that "there is no requirement that a Fedora-trademark-approved web site be run on Fedora - merely that the content be related to Fedora." The issue might have been considered closed there, but Bill Nottingham pointed out that "large chunks" of the Fedora infrastructure run on Red Hat Enterprise Linux (RHEL) as well — not Fedora.
That inspired Jóhann B. Guðmundsson to ask why Fedora
isn't eating its own dog food. It certainly seems a reasonable question
on the surface; why wouldn't a Linux distribution project wish to showcase
the distribution by hosting its infrastructure using its own system?
Guðmundsson goes a bit farther and asks if it "discredits Fedora
as a viable server platform" if the infrastructure runs on RHEL. But
is Fedora really a viable server platform to begin with?
The consensus, unsurprisingly, seems to be "no." Fedora's lifecycle is far too short. Having to upgrade a server, even if a distribution upgrade works flawlessly, every 13 months is simply more work and disruption than most organizations care to face. A 13-month-old desktop may seem ancient to most in the Linux community, but a once a server is configured and doing its job there is often little reason to perform a significant update after only 13 months. And the reality, as Mike McGrath illustrates is that it's impossible to even get the 13 months and not really worth the effort:
So sure, we could run Fedora for everything, especially if we slimmed our offering considerably. But what do we gain? Many of the systems above hardly change as far as their primary purpose goes. Look at our dns servers. They run bind. But they also have several underlying systems that make them part of our infrastructure. Monitoring, configuration management, account systems, etc. So by upgrading dns every 6 or 12 months, literally gain NOTHING. But the time lost in creating the staging tests, upgrading them, running the tests, then scheduling the downtime in production and doing the actual upgrades is actually quite large.
Then there's the matter of Fedora's actual focus as a distribution. Several Fedora contributors have taken the position that it's too focused on desktop usage, and not enough on servers. Adam Williamson says that they're missing the point of Fedora:
To expand on this: remember, a lot of the 'point' of Fedora is to offer a space for experimentation and development. Mike and I are suggesting Fedora is not a good distribution to run boring, stable, important server tasks which just don't change a lot over time on - like DNS. However, if you were prototyping the next generation of DNS server, would Fedora be a great place to do it? Why yes, it would. Would that be partly thanks to the efforts of the server SIG? Definitely. Ditto if you're developing a web server or a web app or a new cloud infrastructure. Maybe Fedora isn't the distro you'll choose to deploy it to paying customers on, two years down the road, but it's certainly a good choice of distro to do your development work on.
To put it another way, one size does not fit all. Some of the qualities that make Fedora interesting for developers and some users make it unsuitable for those seeking a distribution that's usable on "boring, stable" servers. Sometimes boring is a feature. Fedora does have a Server Special Interest Group (SIG), but it has few active participants and seems to be moving a bit slowly. There's no spin from the SIG, though you can find a recently generated Kickstart file.
Can't we have an LTS?
Fedora isn't the only distribution that has this conversation on a
regular basis, of course. It has come up repeatedly in the openSUSE
community, particularly since openSUSE reduced its support
period to 18 months from 24. At the time this decision was
made, I was working with Novell as the openSUSE community manager and
agreed with the decision to cut support — despite knowing that it
would be unpopular.
While not a popular decision, it makes little sense for Red Hat or Novell to fund the support of their respective community releases as long as some community members would like. A few extra months of support are not likely to make a distribution like openSUSE or Fedora competitive with RHEL or SUSE Linux Enterprise Server (SLES) — but they will cost the company a pretty penny over time. One thing that volunteer contributors have not typically been eager or effective at is backporting packages to older releases. The Fedora Legacy project tried, and failed, when it came to supporting Fedora releases past their lifecycle.
In the openSUSE community, discussions of an "openSUSE LTS" or openSLES have surfaced frequently, but often to little effect. Many users are supportive of the idea, but finding the number of hands skilled enough and devoted enough to make it happen is not easy. Sending an email or opening a request is easy. Actually making it happen is difficult — but not impossible.
Wolfgang Rosenauer has been pushing
this idea forward with the Evergreen project, an effort to provide long
term support for openSUSE releases. There does seem to be some momentum
behind the effort and an actual repository
for 11.1 updates in the openSUSE Build Service (OBS). openSUSE 11.1
technically went end of life December
31st, 2010 but Rosenauer noted that the updates hadn't closed yet as of
January 7, 2011. It's unclear where the final repositories will land and
when Evergreen will be able to offer updates by default.
Unfortunately, the archives for the evergreen mailing list are open to
subscribers only — but subscription is open and apparently
unmoderated. At the moment it looks to be a handful of participants with
Rosenauer doing much of the work. He has called for assistance, and noted:
"I've seen basically people offering to test stuff... While that's
important it's not enough to get this project flying. If it would be that
easy the openSUSE Community (or even Novell) would do it
While one wishes success to Rosenauer and the Evergreen contributors,
and to those trying to run the Fedora Server SIG, at best they will create
something that's largely redundant. Organizations that truly need long term
server-suitable distributions can turn to RHEL, SLES, Ubuntu LTS, Debian,
and CentOS, to name a few. There's little real practical need for
repurposing Fedora or openSUSE as infrastructure distributions.
to post comments)