Beta packages within stable releases

Mike McGrath mmcgrath at redhat.com
Sat Oct 10 21:46:38 UTC 2009


How timely.

On Sat, 10 Oct 2009, Ken Chilton wrote:

> Dear members of the Fedora Board,
>

I'm also curious about the !Fedora Board.

> I wish to encourage you to reconsider the policies regarding packages in Fedora releases.  Currently, within Fedora 11,
> there are key packages being distributed which are still in beta or otherwise unstable condition with matching, prior
> stable packages available.  Fedora 11, as best as I can tell, is not a beta or development release.
>
> There used to be a practice in all things Linux where even-numbered release numbers implied stable releases, where
> odd-numbered releases were development tracks.  The Linux kernel follows such a scheme, with 2.4.x and 2.6.x being the
> (more) stable.  I am unaware that Fedora 11 is following this scheme.  Please correct me if I am mistaken.  It also might
> be nice to modify the Fedora Project web page to indicate that release 11 is a development platform and provide clear
> links to Fedora 10.
>

There's a couple of things going on here.  The important ones I think are
the first two of our mission statements:

https://fedoraproject.org/wiki/Overview#Our_Mission

"Always strives to lead, not follow."

"consistently seeks to create, improve, and spread free/libre code and
content."

What you are asking, in my opinion, is to slow down a bit.  This doesn't
necessarily have to conflict with our first mission but it might put it at
risk.  There's been much soul searching related to this very topic over
the last several weeks (I assume you've seen it).


> Two packages of particular note and suitable as exemplars are Firefox and Thunderbird.  These are well-known and basic to
> the Fedora release for most users.  While many Linux developers produce high-quality betas and releases, these two
> packages are worth special attention.
>
> Firefox 3.5.3-1 has both a memory leak and a problem with CPU usage.  When left open for more than a day, with several
> tabs used, the package steadily increases its memory consumption from a few hundred megabytes to over 1.6 gigabytes.  The
> CPU consumption, on a multicore AMD machine, has been observed to start at 40% while minimized to 100% after a few hours
> on non-use.  Users of the latest versions of Firefox have found that frequent killing of the Firefox process and
> restarting is required (this is on Fedora, not Windows).  While this problem has existed to some small degree in the
> past, the latest versions are actually much worse, contrary to the Firefox developers' claims.  While the Firefox
> community continues to struggle with fixes, removal of add-ins, and other attempts to locate the source of the problems
> and placate their users on all platforms, Fedora continues to adopt the latest buggy release of the tool as it is
> unleashed.  It would seem prudent that Fedora have some degree of QA concerning the packages it considers key.  A web
> browser is one of the features that everyone from the mere novice to the staunch professional requires.  Fedora should
> select the best browser available, and not just the most recent or the one with the most features.  It would seem
> appropriate that Fedora should refuse to move forward to newer releases of packages that move backward in quality. 
> Fedora Project should implement its own QA and select the stable releases for its stable releases. This might also be of
> benefit to the Firefox developers, who can spend more time chasing down the problems Fedora has implicated, and less time
> trying to run and tie their shoelaces at the same time.
>
> Thunderbird 3 is currently undergoing many changes.  While developers continue to add more features and new development
> versions are released, those working on the coding and testing of the new features are not at all disturbed by the
> frequent changes to the UI and other characteristics of the tool.  However, those who depend on the email facilities in
> Fedora are likely quite worried when Thunderbird pops up a dire warning about using a beta package for real life.  Anyone
> who would be furious when all of their email, current and filed, is lost because the beta package did what we were told
> it could do.  A user who expects to take a quick check of his email and finds that the whole UI has changed, his
> preferences gone, and previously admissible email now finds a home in the junk mail abyss might be a bit perturbed by the
> advent of Fedora 11.  It would seem quite reasonable that Fedora 11beta would include beta releases, but "Fedora 11
> Release" should have only included Thunderbird 2 in the "release" repository, with Thunderbird 3 in the "testing" one. 
> If during the beta phase of Fedora, a package cannot be deemed stable, it should either be excluded from Fedora or Fedora
> should revert to the prior, stable version of that package.  Beta software is not intended for production environments. 
> Anyone who needs the email to work will not want to rely on a beta package.  Thunderbird 3 has become a black mark on
> Fedora 11 and something I hope the Fedora Project plans never to repeat.
>
> So, I hope this email will be received in a positive light.  I suspect you may have already heard from many others, since
> this seems too big to ignore.  I hope that we might see a change in the Fedora Project to provide stable releases to the
> community while not hampering development.  This might mean adoption of the even/odd scheme, or a more formal QA criteria
> and process, or maybe just slowing down the alpha/beta phases to allow more testing before calling it a release.  I
> believe some of what has happened was in hope for the best, but in the end there must be a right solution and that is the
> one that considers the consumer.  Please remember that the consumer wants quality, not just quantity.
>

I personally prefer to have Thunderbird 3 and Firefox 3.5.3 on my desktop
because I understand when things are and aren't working right and can
communicate that to upstream.  In my opinion if you require older versions
of these critically important software packages, there are other
distributions that offer it.  For example the most recent version of
Ubuntu ships the 3.0.X tree of Firefox.

Others disagree, putting a much higher value on market share and user
quantity.  In that instance, slowing down might make sense.  We would have
a more stable environment to work in, which is desirable by most.  We'd
also be less distinguishable from other distros.  Our recent additional
commitment to QA will help some of the stability issues many complain
about, I don't think there's any plan in place that would detect the
specific issues you are describing.

In my opinion both Firefox and Thunderbird benefit from us doing early
adoption of their software.  Additionally, the more stable versions of the
software packages you are looking to use are even more stable as a result
of Fedora getting them first and helping find and fix those bugs.  There's
a lot of talent here to do that sort of work.

As far as the bigger question of what 'stable' means.  All upstreams have
different opinions on what stable means.  Even if they didn't, our
packagers might disagree.  I know we have guidelines on how to package
pre-release software, I do not know if there are any policies or
guidelines around what is and is not acceptable to be in Fedora.
Generally if it's Free Software and it compiles, it can get in.  I'm not
even sure what body decides this though I'd assume some mixture of the
packaging committee and FESCo.

Looking outside the box a bit, there's also the option of shipping both
versions of some of these 'day to day' packages.  Providing both
additional stability for the people that want it, and a bit of blood for
those that can handle it.  Though this puts additional strain on the
packagers.

	-Mike


More information about the fedora-advisory-board mailing list