Henry Kingman today shares with the Linux.com community his exclusive interview with Fedora Project Leader Paul Frields. Frields goes into detail on the upcoming Fedora 13 release, his decision to transition out of the Project Leader position and how many contributors to Fedora are being paid by Red Hat, among many other topics. Grab a cup of coffee for this in-depth discussion.
- - - - - - - - - - - - - - - -
Kingman: As we approach the release of lucky Fedora 13, how is everything going? In playing with the F13 release candidate, I noticed some updates to the RPM tools; but otherwise, things seem to be pretty stable.
Frields: We're getting toward the end of the process. We've had an incredible response from the community for testing this release. But last night, at a "readiness" meeting, we identified at least one bug that we couldn't clear. So we're now targeting May 25, instead of May 18, as originally planned.
Kingman: As I recall, Fedora usually targets Halloween and International Worker\s Day for releases. So you're actually a little bit late already, right? What exactly did the bug entail?
Frields: What is International Worker's Day?
Kingman: May 1st.
Frields: Oh, right... yes, I usually think of it as "May Day," and hope it isn't. [chuckles] But yes, we originally targeted May 11 with this release, so we had already delayed once. This time, the bug involved booting from an ISO image, and then trying to install from an NFS repository.
We actually found out today that that bug was a little less critical than we thought. If we knew then what we know now, we might not have slipped. But at the time, we felt it was more prudent to wait.
Anyway, [the delay] gains us a day or two to tweak a couple of other minor issues that we found... just little things that might not cause a complete reset of testing and validation. We do have to make sure those minor issues are fixed and that we don't inadvertently induce other problems or regressions.
We're starting to have a very robust and detailed set of release criteria that controls how we decide when to ship.
Kingman: Why do you think you had so much help from the public testing this release?
Frields: Part of it is that we've done a better job providing more capabilities for people to do testing. We're executing better, in terms of letting everyone know, having things ready, and accumulating test results.
We started having test days in the Fedora 11 release cycle, and they've become increasingly popular. For one thing, we've gotten better at providing a live image on the designated day. That guarantees the testing happens on a known platform, across the entire group of testers.
We're also doing a better job announcing test days. We're talking to the open source community about them via the Fedora Planet, blog aggregation, social networking, and Linux.com.
In particular, we got an extremely good response to things that matter to lots of people, like graphics. But we also did test days for storage, the installer, and for particular desktop apps, like scanning and printing, too.
Still, with the test days for graphics adapters, we executed the best, in terms of letting everyone know, having standard live test images ready, and accumulating test results. I'm a pretty busy fellow, and even I found it pretty easy to sit down for an hour and a half, run the test, and post the results to the Wiki.
This kind of broad participation, through test days, leads to much stronger releases. And that's true not just for Fedora, but for everyone, because we work hard to coordinate our changes upstream.
Kingman: I saw some Phoronix benchmarks of the F13 beta that suggested Intel graphics performance had regressed, compared to F12. Do you happen to know if that got fixed?
Frields: I'm not sure. I do know we've had some fixes for Intel graphics driver in last few weeks. Actually, we've made some significant bug fixes for the Radeon and Nouveau drivers, too.
Kingman: Back in March, you announced your intent to pass on the mantle of Fedora Project leadership. Where do things stand currently, vis-a-vis finding someone to take your place?
Frields: The Fedora Project Leader is a paid position, where you work as a Red Hat employee. The objective is to be that one throat to choke, accountable for all things Fedora.
In practice, the majority of time is spent aligning what the entire Fedora community does. That includes those paid by Red Hat, those paid by others to work on Fedora, and the large community of people who work on it because they love it -- the community that any big open source software project attracts.
As the leader of an open source community, one thing I recognize is that healthy turnover is important for the community. It's important at every level, but especially important for the leadership.
During my term, I've had exceptional support from the community and from Red hat. But, I feel like there comes a time when I've done a lot of the things I set out to do. And, there's some evolution and some great improvements left. I want to make sure that someone else can bring their vision, and try to make that into reality.
I've actually spent just a little bit longer than my predecessor, Max Spevack. There's not an absolute number around how long a project leader can stay. In the case of Fedora, two years seems to be about right.
I wasn't encouraged to move on. It's just something you feel it's time for.
So far, I've been able to get us to the phase of identifying a few suitable candidates and assessing who could do a good job. At the end of the day, it's a paid job, so I can't go into detail on the candidates. I am confident that any of those we've identified would make a good leader.
Kingman: Has a date been set for the transition?
Frields: There's not a hard date for the transition. But, I would not expect it to drag into July. We'll probably be bringing someone on during Red Hat's second fiscal quarter.
Kingman: And what's next for you? I assume you'll be staying on at Red Hat?
Frields: I am going to be staying at Red Hat. It's not all set in stone. But my first and most important task is helping the next leader transition into the job -- providing a smooth runway and handoff. I hope to give the next leader all the advantages extended to me by Max. He gave me a lot of support.
As time goes on, I may not be involved in the Fedora community to the same extent. But like Max, I don't ever see myself leaving the community entirely or becoming completely invisible.
Kingman: I guess a change of leadership like that kind of invites some high-level questions. How would you describe Fedora's identity and place within the greater open source software world? Who do you see as the target Fedora user?
Frields: In the context of all other distributions, I believe we have a number of distinguishing qualities. First, I think we do the best job of balancing innovation -- including the latest and greatest open source features -- and also combining that with a completely open and transparent development process. That's truly in support of the open source principles that a lot of people and communities espouse. And, that combo sets Fedora apart, and different effects come out of that.
But if you ask a random Linux user what they'd associate Fedora with, we have a strong identification with "Freedom." That's an incredibly strong thing for us -- to maintain remixability and redistributability. We don't think of our users as just consumers. Everyone is a potential remixer, and we try not to pass along any legal roadblocks. That's why we don't have MP3 or DVD support built into the distribution. Besides patent issues, there's this fundamental issue of distributability.
The second association you'd find is that we're the "grown-up" distro. You might cut your teeth somewhere else, but as you learn more about the engineering and code, Fedora becomes more relevant.
In a business context, of course, Fedora's already relevant. We have a strong link with Red Hat, because Fedora is the upstream source for Red Hat Enterprise Linux (RHEL). So for business users, what they see in Fedora has strong relevance to what they'll see in the future, in their business. Look at the RHEL 6 beta, and you'll see all the features you and I talked about in previous releases of Fedora.
Increasingly, I hear from people at conferences about how they're using Fedora in their families, and in schools, and other areas we may not really have focused on serving. It can start to seem like a lot of disparate targets to hit.
However, that's a bit of a false dichotomy. It turns out that if you solve problems for the general user -- someone who's not necessarily an open source aficionado but just wants to get more out of their computer -- the chances are good you're also solving things for developers and open source aficionados, too. It's important to understand the needs of developers. However, a lot of us do productivity work. There's a large number of factors in common with all of us.
It's difficult to say whether we will ever design for the general user in a way that would target or threaten Microsoft Windows. That would be difficult to do without applying a greater degree of focus and resources. We need to reconcile that with our mission and other objectives, because we do have the ability to provide that focus and do it in an open and transparent way.
Still, there are improvements we can make and are making. The feature list for f13 is pretty impressive.
Kingman: Yes, there are some pretty notable items! Like color management, and a whole new "Design Suite" spin... Macs and PCs have had that since the mid-80s, though arguably driven by third-party commercial providers. Still, it\s rather amazing to think Linux workstations could actually be used in desktop publishing soon, in addition to Internet publishing, where of course they have long ruled supreme.
Frields: Yes, GNOME's color management feature was driven by Richard Hughes, who works for Red Hat. He's also responsible for PackageKit, which we've discussed before and is aimed at smoothing the software management experience.
Richard took it upon himself to work on a color management stack. Apparently, he got in a conversation with some designers, and they noted to him that they had trouble with the lack of a color management solution in Linux. Richard, who could be described as a "good" developer, did the majority of this work over a weekend. Later, he added the capability to import ICC color profiles and other types of profiles.
I tried it out the last time I was in Raleigh, and I was stunned. I had just bought my first digital single-lens reflex camera. Richard had a gizmo you attach via USB, then hang over the screen, and it generates information the color manager can interpret and turn into a profile. After spending a few minutes, I had a color profile that matches the screen on my laptop, one for the DSLR, and one for the printer I have upstairs.
Generating those things is fairly easy to do. You do need some calibration hardware -- a color tile, in the case of the camera, and the USB gizmo, in the case of the display. But when you're done, you have a color work-flow from capture, to display, and output.
We've probably needed that for a while. It took a talented developer to identify the need. He simply identified a need, and designed a system that would fill that gap. A lot of people donated pieces of equipment. That's a way for a lot of people to participate, even though they may not be developers themselves: by sending a few dollars to Richard, who lives in England, so he can buy the things he needed.
Kingman: There seem to be quite a few official new "Spins" of Fedora this release... would you care to comment on any of the others?
Frields: I'm kind of excited about the "Sugar-on-a-Stick" spin. It was driven by a fellow named Sebastian Dziallas. He lives in Europe, but I know him well. He has been involved in the One Laptop per Child (OLPC) project for a while.
As Sugar Labs spun off (from the OLPC project) to work on development of the Sugar environment, Sebastian was involved there. He wanted to be sure there was a complete integrated environment that could produce on-demand builds of a complete OS that shows off the Sugar interface.
Fedora has been the basis for Sugar Software and the OLPC stack for some time now. There was a worry that spinning off the software would result in a disconnect for Sugar. Would releases happen as often? Would the OS provide the platform needed for Sugar? We at Fedora wanted to support Sugar by extending some co-branding between Fedora and Sugar.
With Sugar-on-a-Stick, Sugar can be disconnected from the idea of OLPC hardware. So it can be considered on its own merits, apart from the idea of low-cost, specialized hardware. Any child with a $5 USB trinket has the ability to run this software on whatever machine they have access to.
Kingman: Yes that's pretty neat. And there are a couple of other new Spins, too, like the Haskell tools Spin, and the Security Spin. Lots of Spinning.
Frields: All of these spins represent not just re-mixability, and the extent to which we make our distro open to contributors to bring in software they care about and want to maintain. The spins also represent a way for people who are interested in something off the beaten path to actually gather a community. Spins inherit from the base distribution, but benefits from everything available at http://spins.fedoraproject.org/. So Fedora has its default offering, and the default image that's our flagship product: the Fedora Desktop Edition. That is the most basic functional way for most people to use Fedora. But then you can add components to craft Fedora to be what you want it to be.
For instance, look at Mairin Duffy, our community design team lead. If she travels to a conference full of human computer interaction designers, she can bring a spin that, when run, will provide access to a whole suite of design tools they can use in the pursuits they care about.
Kingman: What about Fedora's support for other architectures, such as ARM? Anything new happening there?
Frields: One change is that PowerPC moved into the "secondary architecture" category. We still provide build software and all the other things. Any secondary arch. community can pick up and build a secondary architecture community.
Several people are working on that for ARM. Chris Tyler at Seneca School of Computer Studies is offering a software building course, where they are putting up a build system for ARM and working their way through.
Kingman: Yeah! I ran across that the other day. It looked like they were using a lot of Sheeva Homeplug devices, which can't have enough memory to support compiling the major apps. But it looked like a fun course. Say, that Fedora build system has a name... what's it called again?
Frields: Oh, yeah... "Koji." It's a kind of yeast used in fermenting Saki, I think.
Another fellow is working on MIPS ports in I believe China -- Gerard Braad. He's a Fedora community member, but officially, Fedora is also interested in strengthening community ties in China. There's an incredible potential community of engineers and computer scientists there to be tapped.
Kingman: Anything else happening in the Fedora Community you'd like to call attention to?
Frields: We have some great events coming up. In July, there's a Fedora user and developers conference (FUDCon) in Santiago, Chile. And, we have a FUDCon event in Latin America in Sept., and in Zurich in the fourth quarter. We also have a FUDCon planned for North America, probably in January of next year.
FUDCons are neat events, focusing on Fedora obviously, but with lots of feature presentations from upstream folks. For example, presentations involving libvirt, and the kernel. We've very much grown up with strong ties to upstream, and that really helps us bring a nice broad view.
That's really a distinguishing factor for Fedora. It doesn't happen by accident, but rather because we coordinate changes upstream. We don't hoard the patches. That helps us avoid this trap of basically putting in a series of hacks to achieve goals not in line with greater open source community.
Kingman: How do Fedora contributors break down, between Red Hat employees, paid contributors from other companies, and community contributors?
Frields: We have an account system that tracks contributors. Something like 70 percent are non-Red Hat, or maybe 80 percent now. These are people who have completed a "contributor" agreement. Among package maintainers, the numbers are similar -- 70 percent or more are not paid by Red Hat.
In fact, I'd be surprised if there are more than a few dozen developers who are paid by Red Hat to work full-time on Fedora. A lot of people come from outside.
Kingman: Oh, something else I wanted to ask... I noticed some changes to the "usual GNOME suspects" in the F13 preview disk Red Hat sent me... like Shotwell instead of F-spot, and GNotes instead of Tomboy. Any reason for removing Mono-based apps?
Frields: We actually removed it from the default spin to save space. We're trying to fit as much as we could into the 700MB CD image. Another change was Simple Scan. XSane has an incredible legacy and is very powerful, but Simple Scan saved space, and is probably also easier to use for the average person. But there were some additions, too, like Pino, for updating Twitter and Indenti.ca.
Kingman: That reminds me... are you still building F13 with the same compiler optimizations for in-order architectures, like Atom, that were used on F12?
Frields: Yes, we retained the Atom optimizations. With every release we look at compiler settings. But we're not dropping 32-bit support any time soon.
Kingman: Well, so many things we didn't even get to, like the new network installation feature. As someone who still tends to install Debian from floppy images, that looks like a neat feature to me.
Frields: Yes, there were several installation improvements, including a much easier to use storage device workpath. And now, people can download an incredibly tiny image and, in theory, use that to install the current version of Fedora over the network, today, tomorrow, or two years from now. I encourage everyone to visit http://boot.fedoraproject.org/ to learn more. As well as the usual http://get.fedoraproject.org/ server.