I hope the title does not mislead. This article is not about how to create your own distribution. These days, we can keep trying a new distribution each day and not run out of them. The problem comes when someone asks which one should he use. We can't even suggest that try them all and then decide! Or even hand them a dozen CD's with 3 or 4 popular ones, and suggest try each and choose.

We make an effort to understand the direction in which GNU/Linux distributions are headed. What differentiates different distributions? A great example of a recent differentiation was the initial theme of a pre-release version of Ubuntu. It shocked enough people that for the official release, this radical theme may not be the default. I felt tempted to try it, but then convinced myself that peace at home is more important.

Each of the major distributions is now very friendly. The range of desktop applications are also very similar. Usage differences lie in the menus and themes. Another area where the differences are greater is in the configuration and administration of the desktop. A new release is no longer that different from the previous ones simply because we have now reached fairly good level of usability of desktops and any significant deviation will inconvenience a lot of users. Is the fun of exploring new distributions and discussing about them in ILUGs over?

We need to look at the next wave. Linspire ( http://www.linspire.com ) has a click and run environment. The application required by a user will be downloaded and installed just when needed. It is very hard for us to visualise the usefulness of this concept in the absence of broadband connections and I do not regard 256 kbps as a broadband connection. Once we are operating in megabits per second, the Internet can truly appear to be an extension of the local disk. With a broadband connection, a GB of mail space on Google may not seem so high after all. Till just a few years ago, we used to regard the downloading of an MB file as arduous. Downloading a few MB files is no longer uncommon. It is reasonable to expect that broadband may be around the corner, at least, in the major metros.

Another intriguing possibility is the Zero Install project ( http://zero-install.sourceforge.net ). “The Zero Install system makes software installation not merely easy, but unnecessary. Users run their applications directly from the Internet from the software author's pages. Caching makes this as fast as running a normal application after the first time, and allows off-line use.” One project which uses this project is the ROX Desktop ( http://rox.sourceforge.net ) . Again, we cannot visualise the possibilities till we get a broadband connection.

Still another idea floating is the componentized distributions. ( http://componentizedlinux.org ) being implemented by Progeny.

Bruce Perens is promoting UserLinux, promoting the idea of a smaller set of core applications because the variety available is becoming is becoming too large for ordinary mortals. Finding ones way around the possibilities is quite tough.

These ideas are all centred around the need to simplify the life for the user. The required applications should show up almost magically on the desktop. They should also get upgraded as and when the need arises.

We should also keep in mind that the distributions are not entirely independent. They are all based on thousands of common projects. Further, most are derived from other distributions. The common starting points are either Redhat/Fedora or Debian. We finally return to the starting question of which one do we use and which should we advise others to use.

Suppose we are dealing with a corporate entity, I can assure you from experience that it is very difficult to upgrade the machines. It is hard to find a time when people are willing to shut down the server for even half a day. If the upgrade causes any problem for a user application, it is chaotic. Hence, the motto is that if it ain't broke, don't fix it. Upgrades and migrations have to be very carefully planned and have to be infrequent. Redhat distribution is aimed at this market. Suse/Novel's distributions will also be after the same space.

Mandrake has also moved to a community edition and an official edition. It is likely that as their installed base increases, the pace of change for the official editions will have to slow down, or the changes will have to be distributed as small, continuous and carefully orchestrated upgrades.

For Corporates, the choice of the distribution will likely be determined by the most appropriate vendor for their business and area of operations.

Conservative home users can choose a similar strategy as the corporates though the importance of the distribution will be determined more by the skills of their PC supplier or a neighbourhood techie. Once set up, any of the common options work very well. Installation of flash, mp3 plugins and mplayer or xine may be hard for Fedora but what matters is the comfort-level of the person who is setting up the home machine.

Now we come to the group which interests us the most – the readers of this magazine. This group will, no doubt, continue to explore and, we hope, build new ones. Knoppix was a brilliant idea, a radical departure from the past and very useful as a new way of using Linux. We now have distributions small enough that they can work off a usb flash disk as well. However, these are not the options for a day-to-day usage on a desktop or a server. For these, my suggestion would be that we work with the distributions which form the base for most of the others, which at present are Fedora and Debian.

Debian is not likely to be a controversial choice. Fedora, unfortunately, is closely associated with Redhat at present. It has been a very good decision to split the non-business segment from the commercial segment and give it the freedom to evolve rapidly and fast. It would be nice to see this project move into the next phase like Eclipse, which was started by IBM but now has a host of organisations supporting that foundation. Xorg is another venture supported by many vendors. It would be good to see other major companies join Fedora or Debian or even both. This will speed up the research and development of new ideas. It will also implicitly lead to greater standardisation across distributions even when the ideas and guidelines are fluid.

We could then have utility programs which will use one these large distributions as a base and create a custom distribution based on capabilities and features desired, including the customised default configuration files. I would have loved to do that when sending software to branch offices. Trying to figure out the problem being faced by Chennai office or Patna office remotely was nightmarish, especially when the problems at each branch are different and not reproducible. For the curious, the problems were usually related to incompatible dll's and virus's.

But my true satisfaction would be to send a CD created just for my parents a thousand miles away, certain that they will not be asked any question they can't answer and the desktop will have only the applications they use with our family photos as the screen savers. Not an impossible dream!