UnitedLinux: Standardising Linux

UnitedLinux, the high-tech version of American Motors, has at least one thing going for it: It's an attempt to begin addressing an old problem with Linux--the inconsistencies between distributions. Without this technical goal, the joint project of Turbolinux and Caldera International in the United States, Conectiva in Brazil, and SuSE in Germany, would barely be of interest.

It's important to bear in mind that UnitedLinux is purely an enterprise play. While the new UnitedLinux distribution will be their in-common product targeted at you folks, the four companies will continue to maintain their own product lines for other markets. Since the Linux enterprise marketplace is dominated by Red Hat, you really have to wonder how big a ripple this will make in the pond.

In fact, a ripple in the pond is a long shot. Of the four UnitedLinux companies, only SuSE has a significant share of the enterprise market, and that market share is concentrated in Europe. In the United States enterprise market, Mandrake and Debian are more significant than SuSE, even if both are small in comparison to Red Hat. UnitedLinux is destined to start out as a bit player.

For many years I've wondered how significant the differences between distributions really were to developers and users. I assume this is really an issue is for developers, since users tend to choose a distribution partly because they prefer the idiosyncrasies of one over another. So to an ISV, do the differences between Red Hat, SuSE, and Debian make a large difference? Do they impose a significant amount of testing and extra development work?

Yes, according to officials at Borland with whom I spoke. As the developers of Kylix--a RAD tool for Linux from the company that invented RAD tools--differences in distributions are a problem for Borland. Officially, they only support Red Hat 7.1, Mandrake 8.0, and SuSE 7.2.

There are a lot of other Linux distributions out there, and it's certainly possible to get Kylix working on them. It's even possible to get these other distributions Kylix-certified since Borland provides a self-certification kit for vendors of the lesser distributions. (Caldera OpenLinux 3.1.1 and ALT Linux Master 2.0 are self-certified in this way.)

So something like UnitedLinux has the potential to make things markedly simpler for an ISV like Borland. When dealing with the myriad distributions in the world today, an ISV must account for different kernel versions and different versions of important code libraries, like the all-important libc or glibc libraries. In some cases, version differences in the gcc compiler used by different distributions can cause binary incompatibilities between programs.

UnitedLinux, on the other hand, will implement a standard called the Linux Standard Base (LSB), which sets uniform rules to address potential incompatibilities. The fact that UnitedLinux implements LSB is, from what I can see, the only point important to ISVs. UnitedLinux would be a lot more important if more than one significant distribution were involved in the effort. As things stand, the amount of testing and development work for Borland probably won't decrease much, if at all.

The other important issues for UnitedLinux are either confusing or unresolved. UnitedLinux will be a single distribution, but exactly what will be included in the distribution has not yet been decided, or at least it hasn't been announced. The UnitedLinux announcement says that "The four partners will each bundle value-added products and services with the UnitedLinux operating system and the resulting offering will be marketed and sold by each of the four partners under their own brands." So, there won't be one UnitedLinux--there will be at least four? I'm confused.

One day, if Red Hat and other major distributions join UnitedLinux--or even if they were to simply pledge to support the LSB--things will be different. That would be a big deal. But Red Hat hasn't signed on, and from a pure self-interest perspective I can't see why it would jeopardize its enterprise dominance by giving numerous competitors an equal opportunity for application certification.

There are some good ideas behind UnitedLinux, and I suppose it's basically a high-road campaign. Darn shame it's unlikely to have much impact on the real world.

What's your take on the UnitedLinux effort? Tell us through Talkback