Creating a Linux Distribution for the Common User

If there were a Linux distribution which appealed to the most common type of computers users, they would be using it already. Some barriers to adoption we can't remove; the fix relates to things we can't control. Yet, in my rant on “rolling release” I tried to point out there is at least one thing we could do differently, if we would — make some effort to support fixing previous releases still in heavy use. That way we can offer something to the one part of the world's computer users we have neglected. T

here are several ways to approach this, and I've received good ideas from people far smarter than me. Since I am loath to harass anyone over buying into my crazy notions about Linux promotion, I have been working on some of these ideas myself. Apparently the craziness is communicable, because at least two developers have contacted me to explore a new Linux distro, or at least modify one (meta-distro) which isn't so far away from our shared vision.

One of the developers is delving deeply into a wide array of projects. These include GoboLinux, Autopackage, and the Rox Desktop. He feels competent to handle it this way, but realizes it will take time. Meanwhile, the other developer is working with me to make some modifications in a current project just so we can test the market, as it were.

Over the past decade, I've tested quite a few Linux distros on a wide variety of hardware. The majority of it was aging hardware, donated to my computer ministry. At any given time, none of the machines were significantly older than what I serviced in people's homes and businesses. On a few occasions, I managed to persuade someone to let me install Linux. Of those, a small portion decided to keep it, once they got used to it. My previous comments about insufficient long-term support reflected the number one complaint of new Linux users from this group of clients — the free upgrades were nice, but kept breaking things. Business clients refused to consider Linux on that basis.

When a couple of Linux companies began answering the need for supporting a system long-term, they naturally had to have a Linux distro which rolled out truly stable releases, and keep them updated with security patches and bug fixes. Currently, the field is led by RedHat and Novell. Canonical is also in the competition. For our project, we knew the commercial offerings of each were were unsuitable, as we intend offering this distro for free. Obviously we needed a way to get those updates in a timely manner without anyone paying a for a support subscription. Currently, Novell does make available source RPMs for SUSE Linux Enterprise Desktop (SLED), but have been unwilling to release updates in the same form for non-paying users. Annual service packs won't get it. In the case of Canonical, there are two issues. While RedHat offers 5-7 year support in the form of source RPMs to anyone willing to build them, Canonical drops non-commercial support at 3 years, and we felt that was too short.

Worse, in the past year or so, anything derived from Debian, including Ubuntu, did not properly configure the X server for most monitors we've tested between us. That is, during the setup process, there is no means to identifying the monitor if the installer can't recognize the signature, so the specs aren't entered in the configuration. On a significant number of monitors still in use among our acquaintances, it required obtaining not just the refresh specs, but proper screen size in pixels, and it all had to be entered manually. For example, Ubuntu and friends consistently failed to recognize my Dell P-1230 20” monitor, leaving me with a mere 800×600 display. That's half the default display mode, and there's no way to correct it using the GUI configuration tools. That's fine for me, since I always keep a current config on hand as backup, but that simply won't do for average non-hobby users. If they can't pull up GUI config tool with a database, and a reasonable chance of finding their monitor model, they'll drop that distro quickly.

In the end, we settled on CentOS, which is directly derived from RedHat's Enterprise Linux (RHEL). We note Scientific Linux is another good choice, but we were more familiar with CentOS. Every derivative of RHEL is well supported by a handful of third-party repositories. This makes it much easier to turn the server-oriented distro into something more suitable to home desktop use. While there is no direct path to some of the multimedia support we all know most home users will demand, we intend to make the path as short as possible. Since the whole point is an interim offering to get the process started, we don't expect to do much more than, at most, simply remaster the basic installer CD. We will remove all the developer and purely server packages, keep one desktop, and add a few known good third-party packages. These are carefully chosen based on a history of support by those third-party packagers.

That one desktop would be the default GNOME 2.16. This has nothing to do with personal preferences, but is simply the lesser of evils for our purpose. It supports the most commonly expected full desktop functions and will remain supported by RHEL, the upstream packager. Should it turn out we can develop a more usable desktop from Rox or something similar, we will quickly make the switch. Frankly, there is no current desktop project which seeks to address the needs of non-techie home users. A major element in our decision process is keeping things as familiar as possible, close to what these users have seen in their past computer experiences. This is the path of least resistance for people who don't consider their computer a hobby, but simply a tool.

Naturally, anyone is welcome to join us. Our fundamental philosphy is simply meeting the mass of ordinary computer users where they are, to the degree possible. This is not about making Linux just like something else, but making Linux easier for non-techies to use. We think Linspire went a long way in this direction, but we want to ignore the profit motive. While we hardly expect such users will ever do anything for Linux in return, the simple strategy of mass adoption should help the make the Internet a better place. I'm sure many of us believe there would be far fewer bot herds, and consequently a great deal less mischief on the Net if more home users ran Linux. While it's possible the crackers could make things as bad for Linux as they have for that other OS, we won't know until we try it.

Ed Hurst is Associate Editor of Open for Business.

<a href="http://lists.rpmforge.net/">http://lists.rpmforge.net/</a>