Older blog entries for mbanck (starting at number 24)

The Ubuntu development model

So, everybody says Ubuntu rocks. I tested it a while ago (and my parents now use it) and I can confirm that. Enough praise has been told about the Ubuntu distribution, so thought I would rather write about the Ubuntu project a bit, as this appears to be much more blurry.

It seems Canonical managed to pull off with a tiny workforce what Debian was not able to do with a thousand volunteers. Of course, there is the mythical man month: about three dozen highly skilled and motivated developers working full time on Ubuntu can somewhat compensate for thousand volunteers of which only a tiny fraction care about releasing at all. However, Ubuntu also bravely decided to take new approaches to distribution development (at least compared to Debian) and try fundamentally different ideas, a couple of which were taken from how the GNOME community works.

The following are the key development point I as an interested outsider gathered from reading their website and from following their mailing lists and IRC channels (corrections/additions welcome):

• They have an infrastructure similar to the Debian one. The archive tools seem to be the same, given that their prime authors work for Canonical and they advertise python as their language of choice. They also have build daemons at least for their three supported architectures. They use mailman for their lists and (for now) bugzilla as their BTS though.
• They split the Debian distribtution into 'main' and 'universe', adding additional packages to the former. The 'main' part is their supported set of Free Software, which comprises their base system, their GNOME desktop and their selected servers and console programs. Universe is basically all other Debian packages which happened to build for them. Next to those, they introduced a 'restricted' component which contains firmware and binary-only drivers.
• At least for their first release, they branched unstable around halfway through their release cycle and worked from there. They concentrate on stabilizing and integrating their main packages and only worry about universe when there is time or interest.
• They rebuild the whole archive (including universe) on their build daemons. If needed, they upload new versions for the packages, sometimes many revisions.
• They have a set of rules that says they should be respectful and communicative between each other. Disputes are regulated by their technical board and community council. This warrants a good working climate between the Ubuntu maintainers, which makes Ubuntu fun to work on.
• There is a rigid, time-based release schedule. Not only the release date itself is fixed well in advance, but also every major milestone along the release process.
• As the release draws near, their release managers have to approve all new upstream versions and later on every fix. They also produce several testing CDs in short time intervals.
• Everybody involved can upload any package (as long as the patch gets approved), there is no concept of NMUs (Non-Maintainer-Uploads). More precisely, there is no concept of package maintenance at all, different developers are just loosely appointed to specific parts of the archive, like X, GNOME, etc. Furthermore, they have teams working together concentrating on specific tasks like server, laptop, security, installer, ports, etc.

Especially the last three points make their release management both more flexible and more rigid at the same time compared to Debian's, and they allow for their strict 6-month release cycle.

It remains to be seen how the Ubuntu development evolves. Some interesting questions in this regard, which will only be answered by time:

• Will all of their work flow back upstream? They modified the standard GNOME desktop, the Debian installation and the project-utopia stack a fair bit, along with numerous smaller tweaks. Whether some or all of those ideas will be proposed for inclusion (and accepted) will be seen.
• Will they seriously try to integrate their distribution-wide modifications into the Custom Debian Distributions framework, or will Ubuntu rather become a distribution framework for Custom Ubuntu Distributions themselves?
• Will they branch off unstable again, or will they just resync the untouched packages? They seem to have modified a huge amount of packages and no central source-repository-management (yet?), so branching/merging might be too cumbersome for them and they might just work from their release.
• Will the Canonical employes who are also Debian Developers continue to dedicate some parts of their free time to work on their non-packaging related tasks for Debian, like account/release/archive management?

My answer to all of the above questions was initially 'yes', and I hope this will continue to be the case. But again, only time can tell.

Of course, it also remains to be seen how Debian development evolves. Ubuntu seems to be the first evolutionary challenge to Debian and it will be interesting to see how Debian adapts to it. It is already clear that Ubuntu will be good for Free Software in general and the Linux desktop in particular, no matter what happens.

I am going to start my Ph.D. next month (well, in case I find a new flat in Munich until then), so I decided to do some Debian hacking while I still have some time. I finally managed to build and upload untampered hurd-i386 xfree86 packages for the first time in history, this should make it much easier to build other packages now. Jeff Bailey said he would restart the Hurd buildd in the near future, so the port should start rolling big time again soon. I also assembled and uploaded a new Hurd package, fixing the permissions problems invoked by the switch to cdbs and adding Marco Gerard's keyboard and mouse repeaters.

Besides that, I also announced a new Debian GNU/Hurd base tarball for xattr-hurd-enabled Linux kernels. Grab the kernel-image or use the kernel-patch, extract the tarball with star, adjust grub and boot into a functional GNU Hurd. It was never easier to get it than now!. The gory details are still here. Somewhat relatedly, Philip Charles has released the K7 series of the Debian GNU/Hurd ISOs.

On the OpenBabel front, I implemented Turbomol support some time ago. I would like to work more on OpenBabel, but Geoff apparently is going to merge the new conversion framework soon, which will result in some shaking. Otherwise, I also cannot get off my ass. I really think OpenBabel should extent to being a base for other chemical applications like ghemical or gchempaint. Those currently use OpenBabel only for import/export of files and then transfer the structure data to their own internal representation. Interaction between the different programs would be much easier if a common ground was used.

I've spent last week in Frankfurt. My parents weren't at home for the most part of it, so I had a pretty relaxing time with my notebook hooked up to our DSL/WLAN router sitting in front of the TV watching the olympic games or outside in our garden, while I abused my parent's P4 box as build machine. After I managed to build GNU Mach with Alfred's NIC-update, I was even able to SSH into it while it ran Debian GNU/Hurd. I finally managed to build glibc for hurd-i386, it took a ext2fs server compiled with a glibc from CVS to circumvent the nasty linkref bug. I also built new binutils and gcc-3.3/3.4 packages, getting the hurd-i386 toolchain back on track. It seems the only thing missing is an update of the hurd package itself, switching the kick-ass Hurd console on by default. Then, the K7 set should be able to go gold.

On Saturday, I passed by the real-life Debian Bug-Squashing-Party, meeting a couple of friends there, most notably Frank Lichtenheld and Peter de Shrijver. The party took place at the Lichtwiese in Darmstadt, the place I passed my first two years of studies in the last 90s. Well, I was not able (or perhaps motivated?) to squash bugs a lot, but just hanging out there during the barbecue was fun enough. Unfortunately, I forgot the AC adapter for my Thinkpad, so I was more or less forced to leave the party at around 2 AM. Many thanks to Martin Zobel-Helas and the others for organizing it.

I also finally did some non-Debian hacking again, adding Turbomole file format support to OpenBabel (CVS commit pending) and hacking on the libghemical autotools setup. It was nice to see that some guys from the University of Iowa are working on ghemical as well, they try to interface it better with the GAMESS software package. However, I believe that support for electron densities and vibrational frequencies should really be included in OpenBabel and hooked into ghemical from there.

So it's been over two months now that I finished my master thesis and am not working at the university anymore. Since then, I did two presentations on the topic of Free Software. One as an introduction to Linux and Free Software in general at my student's association (in german), the other at LinuxTag about the Debian GNU/Hurd port. Both went fairly well, but could have been better.

The rest of the time since early June I spent hanging around mostly. Before LinuxTag, I ported xfce4 to GNU/Hurd, trying to get the port some eye-candy. Afterwards, I mainly concentrated on making sure glibc is fine and updating the xfree86 port for Debian GNU/Hurd, which got finally merged in late July and should be a major leap forward. In early August, I announced a new tarball-based cross-install method for Debian GNU/Hurd which is based on Roland McGrath's xattr-hurd patches for ext2. Unfortunately, it does seem to be too late to get the patches into sarge, either via upstream or via the Debian package. However, Philip Charles has mentioned that he will use this method for the next major round of the Debian GNU/Hurd installation CDs. Elsewhere in the Hurd world, the big news is that Neal Walfied has picked up hacking on the Hurd again and started to review Ognyan's patch for large ext2 partitions (which seems to be getting pretty stable, I compiled xfree86 and glibc several times without major problems). Also, Marco Gerards is still working on integrating the Hurd console and X11 writing a keyboard repeater and he also started looking at DHCP support.

For the non-Hurd related Debian stuff, I mostly picked up my work as Application Manager again (though not as thoroughly as I'd like to) and did some work as sponsor for micah, gravity, ajmitch, jdub (though I managed to mess up his package) and daf. Over the last couple of days, I found the time to get my own packages in shape for sarge... again. Further, I had some more free time to (re)subscribe to a couple of mailing lists like -vote, -project and -devel. However, I decided not to take the pain of subscribing to debian-devel and rather went for the GNOME devel list, which has much more interesting, inspiring and on-topic posts than its Debian counterpart, at least in my opinion.

LinuxTag was a blast. I went there on Wednesday afternoon, quickly walked over the expo, noticed that no GNOME desktop was being displayed at the Novell/SuSE booth and that the Gentoo booth was sort of at the end of the world. I had the expression that the overall expo area had grown since last year, it always took me quite some time to find my way back to the Debian booth. Unfortunately, the hacking area this year was pretty, uhm, limited. On the other hand, the Debian booth was very nice this year. Highlight was a very slick 2m height case where all the merchandise was displayed. The whole booth was very well done as well and looked more professional than last year. Time was quite advanced, so I desperately tried to find somebody to watch the football game (Germany versus Czech Republic) with. Most of the Debian crowd did not seem to be particulary interested, so I decided to stick with the Credativ guys like Noel, Andreas and Michael. However, we walked across half the town looking for some greek restaurant, only to find out it was the wrong one. We finally settled for an italian restaurant just next to the LinuxTag area where I watched the first half of the game. I could not watch the second half though, as I promised to fetch our fearless leader from the Ryanair airport. When we came to the AKK (the gym hall we all slept at), we had a beer and I worked on my talk until around 3 AM. With some guys snoring, I had a hard time falling asleep and I woke up again around 7 AM when the first over-motivated people decided it was time to get ready for another day LinuxTag, sigh.

My talk on Thursday about the Debian GNU/Hurd port went okayish. I delivered it in german, which perhaps blocked me a bit as I had to translate a couple of terms on the fly and had the feeling of repeating myself without need a couple of times. However, interest was pretty high and there were quite a couple of questions, which encouraged me. Unfortunately, magicpoint did not want to run on my Debian GNU/Hurd installation this year and I did not have enough time to debug it, so I had to give the talk on GNU/Linux. After my talk I relaxed a bit and walked around the area. I missed the demonstration against software patents, Martin and me had lunch and went shopping in the city center. In the evening, the famous KaLUG party took place. I have to admit that I liked the setup last year better, at some distance to the AKK, where people sat down in the grass. Instead, the location was right next to the AKK this year, in the same place we used to hang out each night and morning. The people were quite cool though, I had some discussions with Martin, Frank and some others about the general way Debian is heading. I wisely drank a couple of beers this time, so I had less problems falling asleep than the night before.

On Friday I first attended Wolfgang's talk about GNU/Hurd and later Ian Murdock's keynote. I spent the rest of the day hanging around at the front or behind the back of the Debian booth, checking how Huedi, Flo and the others were doing at the hacking contest walking around LinuxTag with Martin and talking to Wolfgang at the FSFEurope booth. I also met Murray Cumming for the first time ever at the GNOME booth (which was right next to the Gentoo booth, I overlooked it on Wednesday) and had a nice conversation with him. He lives very close to me in Munich, so I thought it was funny to meet him at Karlsruhe. Luckily, I could convince Martin to take the Bus back to the airport, so I could attend the social event, which rocked big time this year. Dogi finally arrived while we were waiting for the bus to bring us to the social event location, a lido just outside of Karlsruhe. Apparently, we were lucky to be on one of the first buses, as we did not have to queue up insane amounts of time to get food at the (wonderful) buffet, compared to the people who arrived late. I hang around with the tyrolian Debian section, including Peter, Dogi and Huedi. Later on, I had a long discussion with Andreas about various topics including Gnoppix, and I watched the game France versus Greece together with the others (Funnily enough, I noted later on that the notebook which did the TV presentation was powered by WindowsXP). Some time later on, I bumped into Chris Halls, and we talked a bit about what he's doing in real life and for Debian. I always found it very impressive that he and Rene seem to manage the whole OpenOffice.org stuff in Debian mostly on their own. When I came back to the AKK, I found out that Dogi and the others were still outside, so we emptied the bottle of Bacardi I brought with me, blended with Dogi's orange juice. Florian Lohoff joined us, and told us a couple of hilarious stories on his own, so this was a great night. I guess I was pretty drunk at that point, because I even managed to sleep until close to 9 AM the next morning.

On Saturday I mainly listened to the 'hacking OpenOffice.org' talk by Michael Meeks and partly followed the talk by Georg Greve. I already listened to him last year, so I decided to move on after a couple of minutes (although he is a very inspiring speaker, much more so than RMS in my opinion). In the afternoon, I spent a while talking with Dogi in the beautiful park just outside of the expo area and then walked around the expo for a last time, before I attended the keysigning party However, I have not signed the keys from last years' LinuxTag, but I hope to do better this year. Later on, I discovered very much by accident that Christoph Lameter gave a talk about the performance of embedded systems and I took the oppurtunity to talk to him afterwards and tried to get his opinion on the situation of embedded systems in Debian. While I think that the modularity of the new Debian-Installer and the mainstream advent of Custom Debian Distributions should make for much easier handling of embedded issues, Christoph was not very interested in embedded systems inside of Debian. Rather, he seemed to believe in just sticking a full blown Debian installation on an embedded system (if the resources permit), or otherwise cross-compile something and copy it over. Anyway, I had a nice talk with him and showed him around the Debian booth a bit.

As a summary, I can say LinuxTag was again a nice experience. I only manned the booth for short while this year (as there was not enough space to display and promote the Debian GNU/Hurd port, unfortunately) and rather walked around and talked to people. I met a lot of old friends again and a couple of fine new people, and it is a pity that some others did not show up (mostly mako and Marcus). And of course Joey was too busy for conversation again this year.

I've got a new notebook. Somehow, I managed to get my parents to buy me a Thinkpad R51, after they got me to give my old notebook to my sister so she can write up her Master thesis.

The Thinkpad is extremely slick of course, albeit not extremely small or light. Installing Debian on it was pretty painless, I actually enjoy having a fresh install finally and watching how everything just works. However, it took me a while to figure out how to make exim4 work with GMX and SMTP-Auth and I still haven't figured out how to nicely integrate exim with tsocks, as I am still behind a SOCKS-proxy. Regarding the WLAN, I managed to install the ipw2100 centrino drivers in short time when I finally bothered. It's great to finally see drivers for this available, although they still require firmware.

APM suspend-to-ram is working very well, but I would like to have ACPI to also throttle the CPU and all. Unfortunately, going to ACPI S3 state and back somehow corrupts the WLAN and ethernet drivers, they stop to work after a while. Putting the Thinkpad to sleep and waking it up again makes the drivers work for a bit more. I just hope this gets fixed some day.

Other that that, I'm very happy with my Thinkpad. The keyboard is awesome (although it took me a while to adopt that the Esc key is above F1 now, being a vi junkie) and the display is great. GNOME-2.6 looks more beautiful than ever. Having a state-of-the-art CPU and 512MB RAM also helps a lot when compiling code. The only drawback is that history is repeating, though this time getting the Gigabit NIC working with GNU Mach might be harder. I plan to take the USB-2.0 and FireWire ports into use and buy an external harddisk case so I can take my data more easily with me.

"Sun should do Debian"

That's the private opinion of a couple of Sun people from their blog entries here and here last weekend. Jonathan Schwartz was already quoted in january that Sun would rethink their Linux desktop platform, currently based on SuSE. The question is which platform should Sun adopt instead, and a couple of people seem to push for Debian. The advantages for Sun's JDS would be quite obvious:

• Sun would not depend on another company for their desktop system.
• In contrast to Fedora, Debian is a really stable platform to build on, with a slow and predictible (at least quality-wise) release cycle.
• Sun could gain a significant boost in community acceptance if the adoption of Debian would be done right.
• Debian has always been a good base for others to build on. With the recent boost of development and usability of Custom Debian Distributions, using Debian as a base has become even easier and cleaner. The ideas of Bruce Perens and Ian Murdock are pushing this even further.

But what's in it for Debian? In my opinion, there could be quite a bit, if done right. Companies sometimes seem to be unsure on how they could help Debian. We don't need a lot of money, we've good a pretty good idea on how a Free Software Distribution should work and look like, we've mostly got enough hardware to keep us running (well, thanks to those guys) and we even seem to be getting just enough sponsors to make DebConf rock. But there are areas where we really could need a helping hand. So this is how I think Sun could help Debian:

• They have done a great job helping GNOME's usability efforts. While debian-installer is a huge leap forward in terms of usability from older releases, there is still a lot of possible improvement (let alone for other tools like apt-get/aptitude) and any hint from Sun's professional usability experts would help us tremendously I think.
• Speaking about debian-installer, Sun could perhaps even help in porting it to GTK+/GNOME, something which we plan for after the next release anyway.
• On a lot of 'About' boxes and manuals for the GNOME desktop you find 'Documented by the Sun GNOME documentation team'. Doing documentation is frequently a problem for Free Software projects and there are a lot of loose ends in Debian in this regard as well.
• Sun could also help with thorough testing of the base Debian distribution prior to release. Right now, we mostly rely on end user installation reports, very few developers actually install testing or try upgrading to it. Although we've got a QA team, getting help from professional testers would be very welcome.
• Who knows, perhaps they could even help our sparc port eventually?

All this would of course depend on Sun's level of committment and whether they want to help Debian natively (something Skolelinux is doing with great success) or rather just use Debian as a base and substitute everything that is not there yet with their own (or other's) stuff. The latter would be fine (after all, it's Free Software), but the former would be much more appreciated by the Debian community for sure.

This is all just throwing thoughts together and of course Sun has not made the slightest notion to actually go forth about this officially. But if that should ever happen and if done right this could be profitable (in whatever sense) for both parties and the Free Software Community in general as well. I think Debian should support and welcome Sun in that case, with all due diligence. But in the end, it is our fundamental advantage that we can only profit from other companies' involvements, there is no way to buy us out and then lay off the Hurd freaks afterwards for example.

I've unsubscribed from debian-boot a couple of days ago after tbm started to flood the list with follow-ups on bug reports and other maintenance work. Great job, but too much for me to bear right now, unfortunately. I have to hand in my Diplomarbeit by the end of the month and so far I'm pretty good at avoiding IRC at least. I also pondered unsubscribing from -private and -vote, but things seem to have calmed down a bit lately, so I guess that can wait. Further, I managed to convince our admin at the university to install ion, so now I can use a decent window-manager which should get me a bit more productive as well.

I've met Bdale and a couple of other Debian guys from Munich last night. It was nice to see Bdale again and we had quite some interesting conversations about the current state of Debian and uncle Bdale also told us a couple of wonderful stories from way back then.

On the Hurd porting front, the patches for glibc and vim are in the packager's repositories pending uploads. The python case proved to be a bit harder than I thought but I think I've tackled that one now as well. I finally need to look at Roland's xattr patch for Linux again and check the pending issue, it would be nice to see that one go in some day. In other news, Barry deFreese managed to build the X11 libraries from X.org on GNU/Hurd yesterday which is really promising as X was next on my list anyway.

I also did some work on my own packages, uploading new upstream versions of chemtool and scmxx this week. Already a while ago, I set up a jack-pkg project on alioth with a mailing-list. Now I only have to figure out how to import the debian/-directory into CVS and then tell everybody who might be interested to join the effort. After the shock of the Sarge delay, I was less than motivated to go on fixing RC bugs for a while, and now I've got no time for that anyway. Since then, I've mostly recovered my faith and again hope we will release Sarge soon and that mako will get the FSF to fix the GFDL.

I'm a bit worried about the firmware issue. I support aj's decision to remove GPL'd firmware without obvious source availability, but I believe we should coordinate this with the other distributions to perhaps put a little more pressure on upstream or the vendors to fix this. After all, this is an issue for the whole Free Software Community. I tried to raise the question on slashdot, but my story did not hit the front page. The comments were mostly rather in our favour though, with the usual threshold of flamebaits and trolls. So this makes me hope we'd have the support of the wider community.

Even more worrying with respect to the kernel and releasing sarge than the firmware problem is the kernel security nightmare. I wonder whether we've learned the lesson from the woody release or whether we'll get a little bit of history repeating:

<Joey> aj: We cannot support both potato and woody
<Joey> supporting 17 architectures... I don't want to think about that...
<aj> Joey: so does that mean you'll be supporting none, or a select "n"? if the latter, which?
<Joey> aj: The current answer would be: no potato updates, and no woody updates, probably
-- #debian-devel, Mon Apr 29 2002 (two days before the designated release date)

At least, we seem to have identified the problem a bit earlier than last time and we don't need to change our archive infrastructure again this time.

So, what have I been up to? Over the last week I managed to get vim and python2.3 packages built for hurd-i386. Both being a bit hacked, they are on ftp.gnuab.org for now, but it seems that they could go into the main archive soon. Right now, I'm trying to build packages of the last big pile of rotten bits left in our toolchain, libc0.3. While I was worried that the GNU Hurd development seemed to be stalled again after a big boost at the beginning of the year, this is fortunately not the case. A new CVS module called 'fabrica' was created, which will contain the designated device-driver-framework for the Hurd on L4. Also, to my utter astonishment, Thomas Bushnell, BSG has started a discussion on bug-hurd all by himself, apparently inspired by reading code.

To my surprise, I really managed to stay off IRC for the most part of last week, which gave me a bit of extra time to fix one RC bug per day. Let's see whether I'll keep this up next week as well, after all:

<mhelas> bugsquashing is like drugs. if you started once, you will never give it up anymore.... ;-)

NP: GangStarr - Put up or shut up

So the hacking party is over. Besides Jens Schmalzing and me, most notably Frank Lichtenheld (the main BSP coordinator, coming from Karlsruhe), Andreas Barth (from the Debian QA team), Norbert Tretkowski (of backports.org, coming from Nuremberg) and Sven Luther (debian-installer powerpc port, coming from Strasbourg) were attending. Frank already arrived on friday and we both hacked at my place that evening. The main event began on saturday and after setting things up we were busy squashing bugs. We all went to a Biergarten for dinner and then continued hacking until 4:30 AM, which seems like a good indicator that poeple were enjoying it. After a very minor amount of sleep, we continued on sunday. The wheather was really excellent then, so we again went to a Biergarten in the afternoon.

The general mood was really good and people said they'd liked it quite a lot, which was also my personal impression. Coming together physically seems to be a good idea as we got a good number of bugs fixed or pending, while the only major contributions from people not in Munich were done by Goswin von Brederlow and also Micah Anderson. I wonder whether the other Debian developers don't feel like fixing bugs right now or perhaps are too busy or something. In addition to most of us fixing bugs, Jens told me that and he and Sven made quite some progress on the powerpc kernel front.

Personally, I also collected various patches to annoying bugs in dpkg on hurd-i386 making a package out of them for testing and managed to port the vim package to Debian GNU/Hurd for the most part, although it still needs some cleanup. I intend to take a break from Debian (or rather, IRC for a start) for a while now because I need to concentrate on finishing my Master Thesis until the end of May, unfortunately.