PCQ Linux 2004 Project New Dawn1

Normally, discussions on another release of PCQLinux start on a lazy note, during a tea break: Should we, should we not? After all, there are enough commercial distros around. Is it worth the trouble? Will our readers want a new one now? What is the release schedule of the major packages, the kernel? The questions have now become routine....

Over the next two weeks or so, all those who want to be involved are expected to keep their eyes glued on to reader mail and also pop the question to anyone they happen to be speaking to: Friends, colleagues, clients, users....

The next step is also fairly predictable. All of us invariably meet again, with time to spare for a more detailed discussion. Normally, one out of two such discussions leads to work starting on a new distribution. Work starting is a fairly loose term. Basically, a team leader is appointed: This time and the last, it has been Vinod. Somebody else works out a schedule: This time it was Anil, previously it used to be me.

The team sort of falls into place automatically. And then the discussions begin to get heated up. What should we base it on? What should we call it? What all should be included? One major point of debate every year is how many CDs? Normally, the choice is between one or two. But falling CD prices gave me the luxury of cutting short this debate by making a grand announcement that we can afford three. numbers.>

This time there was a twist to the tale at this point. Normally, the debate on which distro to base on is a straight fight between RedHat and Debian, with RedHat winning so far. But this time, there is no RedHat in the ring. That opened up the debate and Kishore pitched in with a strong case for Debian. But, at the end, the majority went with Fedora. Going by the trend, I have a sneaking feeling that the next time around, Debian will have a stronger case, and may even win!

That brought us to the name bit. We needed two names, the distro name and the project name. Should it be PCQLinux 9? Or, something else? It was Shekhar's argument that won the day. We named the distro PCQLinux 2004 and not PCQLinux 9, to signify the move away from RedHat Linux. The project name was my choice: NewDawn. Either the others liked it, or were too tired to argue, and wanted to get on with the job at hand. Why NewDawn? Please read the adjacent box.

Next came a big experiment. We wanted to know what you wanted to see included in the distro. Earlier, we had relied on e-mail. Now we took the step of making it more public. We set up an online forum where readers could post what they wanted. The response was amazing (see box). We have tried our best to include all that we could. But, unfortunately, all suggestions could not make it to the final release.

Now started the hard work. The time to get the hands dirty, the hands sweating and the test machines crashing. 

Here is what happened in Vinod's own words:

“We started with the tedious task of sorting, understanding, removing and testing dependencies of all the packages we wished to remove. Language packs, as usual,were the easiest to do. Next came the ones that were duplicate, or the ones that we thought we could give better alternatives for. 

As new software was being added, we also started working on customizing anaconda, the installer. Problem! We couldn't figure out why the source of the installer didn't contain any reference to Fedora, until we realized that Fedora, for some strange reasons, had supplied a generic anaconda. That meant that the anaconda source provided would not compile to what was installing Fedora. Further, the installer code had no proper structure, hardly any comments within the code, and as usual zero documentation. In fact, on the anaconda developers' mailing list, the lead programmer admits that the structure and coding of the program is known only to him and he has no written notes. (That leads to a big question on the openness of such projects and the need for better processes and documentation all around. But that is for another day.)

Anyway, we figured out most of the code over multiple sessions of debugging and tracing through (by going back to the anaconda we had created for PCQLinux 8) and finally came up with a working installer. 

Sanjay, our downloader and tester-in-chief, had meanwhile gone on long leave to get married. And then came LinuxQuest 2004, which kept most of us busy and out of station. However, thanks go to Anindya who volunteered to stay back and slog alone at nights.”

Bendi, the PCQuest designer, was meanwhile roped in to do the graphics and colors for the distro. His excellent work is what you get to see on your screen during installation.

One big event we were waiting for was KDE 3.2. Sirtaj Kang, a KDE developer, was with us for LinuxQuest. But, he refused to help us with any date predictions. We were almost sure that we will have to give 3.2 a miss, when it finally got released two days before we were to go into manufacture. A mad scramble had us download, test and replace, all in a single day. 

Meanwhile, Shekhar, our Oscar expert, was toiling away to get Oscar to run on Fedora. I remember him working away on his notebook, even at airport lounges to get Oscar ready on time. But, unfortunately, luck deserted us on Oscar, and we could not include a working Oscar with this distribution. Hopefully, we will have it in one of the coming months.

Another headache was carrying two kernels on the same CD set and getting 2.4 to install first and then upgrade to 2.6. Two files of the same name, differing only in versions, to be carried on the same CD!! But, we had promised 2.6 to our readers! And, carrying 2.6 alone is risky as it is still experimental!!! Finally, Vinod and Anindya figured out that one, too. It is a little bit of “if” check in the rc.sysinit file, that is down every time the system boots. The same method was then extended to install openMosix and Kernel NTFS patch.

Sun gave us permission to distribute JDK 1.4.2. But, that was a bin and not an rpm. The bin on execution shows the license and on your acceptance, extracts the rpm. It required some deft workarounds to get it to install smoothly. To give away the trick, it is not installed from anaconda, but from rc.sysint, with the license and packaging intact. Whew!

One day to go for manufacture and no major emergencies. Too good to be true? Yes, it was. Kernel 2.6.2 RPMs came out just as we were going to create the masters. But, by now we were used to this routine, and knew what to do. Scramble, scramble.... And were through!

Or, so we thought. Till a major bug was found in comps.rpm, after the CDs had shipped! Again, we were used to that. Call courier. Stop CD. Change comps.rpm. Recompile, remaster, test, test, test, recreate master, pack, and label for dispatch are all but a night's work!

Easter egg: Find the PCQLinux team that built this distribution somewhere in there and be the first to post it on the forum! Have fun with the new release.

Krishna Kumar

LinuxForum

In another first for a publication in India, we ran an online forum (www.pcquest.com/content/linuxforum) inviting ideas and comments from our readers about what to include in PCQLinux 2004.

Starting off on 2003 Christmas night, the Forum ran for about two months, attracting 281 original posts and 190 follow-ups. Many software that have been included in PCQLinux 2004 are based on suggestions at the Forum. We would otherwise have missed them.

Arin Basu (suggested R-Statistical system), Vijay Patil (DigiKam), G Karunakar (Indian language keymaps), Tirukumaran (Eclipse), Deepak Kashyap (SCILab), Hunter and Yogesh are just some of those who helped make this distribution as good as it has become.

I would like to thank all of you for your contributions. Based on the success of this experiment, we are now expanding the Forum to other areas. The first of these is to provide support to this distribution itself.

Support will be provided at the same URL. May we also look forward to those using the distribution to volunteer to help the others?