Fedora QA (Qualety Ashurunce)

Technorati Profile

Today was a sad day for the Fedora project. Some may say that I’m making a mountain out of a mole hill, others would say storm in a tea cup. No-one would say storm on a mole hill, or mountain from a tea cup. I abandoned Fedora when it was at its core level (Fedora Core 6) for a brief stint with the “in” crowd at Ubuntu HQ. Very shortly after, upon realising that I looked like the identical twin of the other “cool” guys at the party, I jetted over to the Novell gang for a bit of opensuse love. A Very configurable distro, and YasT (now Yast2) is a sturdy piece of work, although I found use for a small handful of the tools it offered. Things turned sour for me with the inexcusable handshake with Microsoft, and things have never been the same again – and I’ve been looking to move distro ever since. Then skipped along the ring road to find myself back at Fedora 10.

Marginally better than Ubuntu

I’ve always had a soft spot for Fedora, even during my time away. I would spend more time tracking “feature lists” for Fedora than I would for my own distro. I feel that Fedora 7-10 has been an inspiring journey for those who’ve held on for the ride. My general feeling is that as the competition to grab users is intense, it’s often quite difficult to sway people away from their current environment, even if the distro is 2% better than the next best. I feel that Fedora  could be at that stage, i.e. marginally better than Ubuntu et al. This puts it in a position where should a user dislike some very minor snag or feature in their current distro, Fedora 10+ may be the one they choose. That’s the quality of distro’s these days – people change their distro of choice when it screws up, rather than a more proactive approach to jumping ship.

So here I am. Running kde 4.1.3 on Fedora 10. Until today, I’ve been loving it. I love the Plymouth project, and looking at “ubuntu brainstorm”, it looks that it’s a technology that will be adopted by other distro’s. I also love the way Fedora has driven the changes in pulseaudio and the webcam (gspca) integration in the most recent kernel. There are tons of other stuff.

Now for the bad news. One thing that annoyed me about Fedora 6- was the constant struggle with Yum, the Yum GUI and failed rpm dependencies. It was very off putting. However Fedora has come a long way right? The Distro controls are very mature now? They’ve established a 6 month cycle, so that users will see a “snapshot” of the Distro software cycles at it’s most reliable and stable state. Quality assurance to ensure that everything is fully tested and that the software it releases (especially the distro specific software!) is 100% stable, 100% of the time.

Bad news

Today Fedora showed me that, for them at least, this is not the case. I point your attention to Packagekit – the up and coming package management tool. Some would say that it’s already here. The Fedora folk think so, and it’s the default software package manager by default in Fedora 10. I’ve been running Fedora 10 for some weeks, and everything’s just fine. I received a Fedora update, from their official repositories, for “bugfix” PackageKit_0.3.11-4.fc10. The bad news – it completely broke my PackageKit software. Now when I try to launch PackageKit I get:
kpackagekit-0.3.1-4.fc10.x86_64 from installed has depsolving problems
–> Missing Dependency: libpackagekit-qt.so.10()(64bit) is needed by package kpackagekit-0.3.1-4.fc10.x86_64 (installed)
Error: Missing Dependency: libpackagekit-qt.so.10()(64bit) is needed by package kpackagekit-0.3.1-4.fc10.x86_64 (installed)

The details of this “bugfix”, or “Package manager destroyer” can be found here.
The maintainer of the software for Fedora, Richard Hughes accepts that it is a problem in his blog.

Since my experiences, a patch for this has been released, version 0.3.12-1.fc10. However, those people who downloaded the earlier “bugfix” will have difficulty even using their version of Package Manager to upgrade to the fixed version!

Systematic failure

It begs the question – where was the Fedora Quality Assurance on such a potentially important package (should it break). What’s the point of having package status – testing, pending, stable etc… if they are not honoured by their title. It doesn’t say much for the Linux reputation if a show stopper can so easily be leaked into a “live” distro. I am in no way blaming Richard Hughes – everyone makes mistakes. I’m saying that there should be no way for a “package release mechanism” to allow through such a blatant flaw. Was there any testing on this package? There can’t have been? surely! Otherwise the tester would be in a position that I’m in now. I will update to the fixed version via the command line later tonight.

There is a misconception (I would hope) from the outside looking into the Linux community, that Fedora is the test bed for Redhat, opensuse the test bed for Novell etc.. etc.. We, in the linux community would argue strongly against that for various reasons. But we have to accept that these sorts of incidents do nothing to help our argument.

For some people, this glitch only a few weeks after the “Fedora 10″ distro was released, will be enough to jump ship. As I mentioned earlier, the competition amongst Linux distro’s is rife (which is a good thing!), that it doesn’t take much for people to sway away, and don’t expect them back any time soon.

In software development, you can only ever achieve 2 out of the following 3 targets: Speed, reliability and innovation. If you want reliable innovation, it will take some time; if you want reliable software written quickly, then it won’t be very innovative; and finally, if you want to produce innovative software quickly, it won’t be reliable. In the ideal Linux world, we should never opt for the final option. Instead, the decision on software management should appear as a pendulum. We should nail the top of the pendulum to the top of that triangle – onto the “Reliable” angle. There should be no way that our software should fall away from that feature. So next we have a choice – do we opt for a quick release, or do we opt for a more substantial and impressive software realise, which takes time to produce.

Today I saw one software release from the release team at Fedora that had fallen away from the “reliable” angle in our triangle of software targets. This should have never had happened. It’s a systematic failure, and if the current “testing approach” is taken by the Fedora team, who’s to say a similar break won’t happen again (and again). I would like to see some sort acknowledgement from the FESCO team on this, as I’m sure 1,000’s of people were affected.

And finally, I am a massive fan of open source software, and Linux. But what’s the harm in self criticism? If we can see our weaknesses, then the future looks bright. There’s a company I know of that never admits to mistakes. As a consequence, they now make “adequate at best” software, and innovation is not in their vocabulary. The company even makes operating systems apparently, though their releases cycles tend towards the 5 year mark.

I’d rather be on this side of the fence.