Achieving a Linux standard LinuxWorld.com 2/16/01 Nicholas Petreley, LinuxWorld.com Even though Mandrake boasts that its Linux distribution is the hottest in retail sales, the company considers its retail product a loss leader. SuSE laid off two-thirds of its US employees. TurboLinux is cutting back on its work force and may soon wed Linuxcare in order to refocus its efforts on services. Stormix, a Debian-based commercial distributor, recently filed for bankruptcy. Corel is getting nowhere with its Linux distribution. In other words, most Linux distributions -- even the ones with growing market shares -- are coming to the conclusion that they can't make their money selling Linux.

Advertisement On this topic ITworld.com Today. Sign up Now! ITworld.com Product Spotlight. Sign up Now! Enrolling in Qt University How a Linux standard would benefit distributors What happens when you throw the switch?

Welcome to the reality of selling a free operating system.

There are a number of reasons why Linux distributions are dying or shrinking and, I predict, will soon consolidate. The first and foremost is that most Linux distributors continue to base their added value on things like ease of installation or additional packages on the CD.

LinuxWorld.com links LinuxWorld.com home Best of LinuxWorld.com The Legacy Files The Penguin Brief Version Control Linux links Linux forums

Once upon a time those were among the primary selling points of a Linux distribution. But now that Linux is becoming widely adopted, the focus has changed. Businesses care less about ease of installation and more about whether their choice of Linux installs and runs at all on their hardware. In many cases they are simply buying server hardware with Linux preinstalled.

And when it comes to software packages, businesses care less about what comes on the CD-ROM and more about whether they can install software that is available via the Internet or from independent software vendors.

Ease of installation and added packages just don't cut it anymore. The biggest shame is that some distributions already have a lot of added value such as specialized software, training, and services to offer today's businesses. But vendors can't market those advantages properly because their base systems are incompatible with the de facto standard, Red Hat. If incompatibility weren't an issue, customers would have an easier time noticing distributions' value.

The issue of installation in particular haunts almost all distributions except Red Hat, because Red Hat is the de facto standard. To some extent Mandrake is immune, because it is largely Red Hat-compatible. But users of other distributions are all-too-familiar with the problems of installing software from the Internet. Just because Caldera, SuSE, TurboLinux, and others are, like Red Hat, RPM-based distributions doesn't mean you can expect to install any RPM-based software package without difficulty. You can't.

I've mentioned that problem a number of times in my columns, since it is one of my biggest pet peeves. When you install an RPM package, it doesn't check the software you actually have on your Linux system to see if the package meets dependencies. Instead, it checks the RPM database. So suppose you happened to compile your own OpenSSL libraries. Or perhaps you somehow manage to install OpenSSL libraries using an RPM file from a different distribution. In the former case, you wouldn't add an entry to your RPM database that tells RPM you have the OpenSSL libraries installed. In the latter case, your RPM file from the competing distribution may create an RPM database entry that does not contain the information that your favorite distribution expects. When you install an RPM package from your current distribution that depends on OpenSSL, it will complain that it can't find the OpenSSL libraries, even though you know you have the OpenSSL libraries installed.

Most people get around problems like that by installing their RPM package with switches such as --nodeps and --force. Anyone who has done that enough knows that sooner or later you're going to break your Linux distribution in such a way that the mess could take hours to untangle.

Scott McNeil, former SuSE president and current open source strategist for VA Linux, founded a company called Zenguin to solve exactly that problem. Zenguin would have created a universal installer that resolved dependencies across distributions and even package formats. But Zenguin dissolved before the product was built. I think there is still a need for such a product, although eliminating the important differences between distributions must actually solve many installation problems.

Linux Standard Base (LSB) will be the first step toward eliminating differences among distributions. Coincidentally (or not), the same Scott McNeil that started Zenguin was recently appointed executive director of the Free Standards Group (LSB's mother organization). That gives me more confidence that we will see an LSB standard emerge in the very near future. McNeil was my proposed candidate to be the executive director of LSB but I'm quite pleased to see him take an even better position.

But even though LSB should finally get off the ground, the distributions still need to agree upon a much more comprehensive standard. It makes no sense for dozens of distribution vendors to duplicate efforts and create products that perpetuate the incompatibilities that hinder their growth. If instead they all agreed to start with a comprehensive Linux base distribution and add value from there, they could spend their time working on added value that means something to today's business Linux customers. Vendors would have a realistic chance to compete based on their added value because today's compatibility problems would no longer stunt their sales.

The only company that probably won't eagerly endorse such a standard is Red Hat, because it is currently the de facto standard distribution. But if all the other major distributions put together a coalition quickly, there is still time to put a lot of pressure on Red Hat to play ball. The other distributions combined have a larger worldwide market share than does Red Hat alone. If they can agree on a standard and get it out the door quickly, then the new coalition standard would become the de facto standard, and Red Hat would suddenly become the incompatible distribution.

Red Hat would inevitably hop on board. At that point, all Linux vendors could stop wasting time managing the base distribution and instead focus on developing and marketing added value in the form of high-level software and services.

Now, if you think it's a radical idea to suggest that all commercial Linux vendors should build their products on a single comprehensive standard distribution instead of differentiating on things like installation and packages, just wait until you find out how I believe that new standard distribution should be created. I'll drop that bomb next week.

Nick Petreley is the founding editor of VarLinux.org (www.varlinux.org). Reach him at nicholas@petreley.com.