Linux Core Consortium: Déjà Vu All Over Again

by Timothy Prickett Morgan

If there is one thing that makes the Linux community and most of the vendors that support it nervous, it is not competition with Unix, Windows, or other platforms; it is not that Linus Torvalds will take a job with Microsoft; and it is not that the SCO Group will prevail in its attempt to undermine Linux sales as it protects what it alleges is its Unix intellectual property. No, the one fear is that Linux will fork and then splinter, just as commercial Unix did more than two decades ago. This fear is the main reason that the Linux Core Consortium was announced recently.

The Linux Core Consortium is backed by four relatively minor commercial Linux players: Brazil's Conectiva, France's Mandrakesoft, Japan's Turbolinux, and America's Progeny. The aim to implement the Linux Standards Base 2.0 specification jointly as a shared set of source and binaries that they will all distribute as the heart of their own Linux distributions.

Linux industry juggernauts Red Hat and Novell have expressed support for the LSB standard, which is created by the Free Standards Group, and for the LCC effort. But neither company said that they would actually participate in the joint development effort the four smaller Linux players are taking on, and they certainly have not said that they will distribute the resulting LCC code as the basis of their respective Red Hat Enterprise Linux and SuSE Linux Enterprise Server distributions.

While LCC is admirable, it seems unlikely that either Red Hat or Novell will actively participate, since mucking about in the Linux kernel and implementing enhancements ahead of each other is how Red Hat and Novell are differentiating themselves.

The LCC announcement will cause many in the IT industry to have a sense of déjà vu; depending on how old you are and/or how much of the brown acid you took at Berkeley, some of you will have multiple flashbacks.

In the summer of 2002, years after Red Hat had gone public and leapt ahead of the other commercial Linux distributors in terms of marketshare and mindshare, SuSE, Conectiva, Turbolinux, and Caldera Systems formed an alliance called the UnitedLinux Consortium that sought to deliver a single, unified Linux code base that these four vendors would co-develop and would ship as the heart of their respective Linuxes.

Unfortunately for the UnitedLinux effort, Caldera Systems decided to buy X86 Unix software vendor Santa Cruz Operation in May 2001 to bolster its position in the market. In August 2002, Caldera renamed itself as the SCO Group and shortly thereafter came to the conclusion that Linux contained stolen Unix intellectual property that its newly acquired SCO unit owned. Caldera, which is a spinout from Novell (which sold many of the rights to Unix to SCO years earlier), then sued IBM in March 2003, killed its Linux business, and pulled out of the UnitedLinux effort. UnitedLinux 1.0 made it to market behind SuSE Linux Enterprise Server 8.0 and similar releases from Conectiva and Turbolinux, and then it just went away.

The Linux Standards Base specification was launched a few months prior to the UnitedLinux effort, and it is really all that survives in terms of cooperation among Linux distros. LSB 2.0 was released at the end of August, and all of the major server vendors and Linux distros endorsed the standard, which seeks to enable application compatibility across Linuxes running on a wide variety of platforms. It doesn't ensure that all Linuxes are exactly the same, but rather that all applications that are written for LSB compliant hardware and software will run on any other LSB compliant system. There is still plenty of room for differentiation in the Linux racket. Maybe LSB can't keep Linux from forking, but it is trying to maintain application compatibility across different Linuxes, and that is something. LCC will take this process one step further and actually create a unified Linux that is shared by the four vendors who are sponsoring the work; perhaps others will join, but this seems unlikely.

No matter what, the variations on the Linux theme are nowhere near as diverse as was the case with Unix--at least not so far. The Unix world has sustained not one, but many forks. AT&T's Unix System V was the foundation of the University of California at Berkeley's Unix project, which Sun Microsystems founder Bill Joy turned into the commercial Berkeley Systems Design variant of Unix and eventually Solaris. The university and AT&T's Unix System Labs battled each other in a landmark court case, and seemingly settled the matter enough that in 1987, Sun could take its variant of BSD (then called SunOS) and reunite it with the AT&T Unix to create Unix System V Release 4. At that point, you would have thought Unix vendors would have seen the necessity of uniting Unix, but no, the Sun-AT&T effort, which created the Unix International standard, actually spawned the opposing Open Software Foundation/1 variant of Unix, which IBM, Digital, Hewlett-Packard, and others decided to back. OSF/1 was never a product, Sun ended up with its own fusion of BSD and SVR4 (which we know as Solaris), AT&T sold Unix off to Novell, and none of the vendors in the OSF besides DEC ever shipped a product that adhered to that spec (that's Tru64 Unix, and it is essentially dead). To further complicate things, there are three variants of the open source BSDs--NetBSD, FreeBSD, and OpenBSD--and two variants of SCO's Unix, OpenServer and UnixWare.

If history is any guide, an effort to unify Linux factions should spawn an opposing faction. Luckily, at this point, there are two opposing factions: Red Hat and Novell.

The LCC members hope to have the common core Linux finished in the first quarter of 2005 and will put LCC releases on a cycle of a new release every 18 to 24 months, thereby providing the kind of stability that enterprise customers require. The first LCC release will be used as the basis for Conectiva Enterprise Server, Mandrakesoft Corporate Server, Progeny Componentized Server, and Turbolinux Enterprise Server. LCC is encouraging other companies and interested parties in the open source community to join up. They plan to implement LSB 2.0 with the extensions that are being worked out right now in the LSB futures group and they also plan in future releases to include guidelines developed by the Open Source Development Labs working groups. (OSDL is where Linus Torvalds gets a paycheck these days.) LCC is supporting 32-bit and 64-bit X86 architectures from Intel and Advanced Micro Devices (Xeon and Opteron, practically speaking) as well as the 64-bit Itanium architecture. (IBM would be wise to get support for the Power platform worked into the project.) LCC says that over time, it wants to develop a common Linux binary core that can be the core of Debian Linux as well as the RPM-style Linuxes typified by the alternatives in the market.