07 Sep 2003

At Last, the Evidence

Last month we finally learned what all the noise is about. Although SCO continues its game of hide and seek with the truth, at least a few of the critical pieces have fallen into place. Naturally, as one would expect given the disingenuous build-up, what we can see of the SCO evidence of copyright infringement in the Linux kernel is almost the square root of nothing.

SCO, in its public statements, has made two sets of claims about supposed copyright infringement by the Linux kernel. In the first place, SCO has claimed that a small number of lines of code (usually described as “about 80,”) can be shown to be literally copied from “Unix.” Sometimes it has been suggested that this is merely an example of a larger set of literally copied lines, but the number has been frequently repeated. More recently, SCO’s CEO Darl McBride has spoken of “hundreds of thousands” and even “a million” lines of code in recent Linux kernels that result from “derivative works” of Sys V Unix.

In August, thanks to Mr McBride’s showmanship, the public learned some of the facts for the first time. During a speech in Las Vegas, Mr McBride projected the lines of code that constitute the only example of literal copying so far alleged. A single routine, for memory allocation, which has been around in various Unices since long before Sys V, and has been published more than once in books on Unix internals. The particular code Mr McBride showed was, moreover, released by SCO’s predecessor, Caldera, under the BSD license, permitting free copying, modification and redistribution by anyone.

SCO has one small point, however. The version of the code released under BSD by Caldera had a Caldera copyright notice on it. Someone at Silicon Graphics took that code and included it in a port of the Linux kernel for SGI hardware, and in doing so dropped the copyright notice. Preservation of the notice was required by the BSD license, so the distribution of the code involved was not compliant with BSD. Whether, given the many previous publications, the elided notice was accurate, and the relevant code was actually copyright Caldera is a separate question.

So, assuming that the removed notice constituted a BSD license violation, how many running Linux kernels around the world are tainted by this infringing code inadvertently created at Silicon Graphics? Zero. The code that SCO discovered was placed in the kernel port for SGI’s IA-64 implementation, which is hardware that SGI never shipped. The IA-64 architecture, otherwise known as Trillium and Itanium 2, is currently represented in the market by HP hardware products; the SGI code is unnecessary, and HP has replaced it in the current IA-64 branch of the Linux kernel. More importantly, Itanium 2 hardware is not exactly common. The critical point is that no PC system anywhere in the world, regardless of which version of the Linux kernel it is running, has ever had this tiny snippet of controversial code in it. Despite all the flashes and bangs, SCO has no claim against any developer, distributor, or user of the Linux kernel or other free software for Intel-architecture computers based on this code.

What of the supposed “hundreds of thousands of lines” of “derivative works”? Mr McBride has said in several interviews that he means code from Unix variants written by and copyrighted by third parties (such as IBM’s AIX), over which Mr McBride claims that SCO has “control rights.” Whatever these statements mean, they don’t mean that parts of such works, contributed to the Linux kernel by the copright-holder and released under GPL, are “derivative works” of SCO’s code in a copyright sense. Under US copyright law, a derivative work means not any work “derived” from prior works, but one which has “substantially copied” from a prior work. In other words, in order to show that code from AIX contributed to the Linux kernel by IBM somehow infringes SCO’s copyright as an unlicensed derivative work, SCO would have to show that there is substantial literal copying from a SCO copyrighted work into AIX and thence into the Linux kernel. No literal copying, no copyright infringement. No copyright infringement, no right to demand license fees from end users. It’s as simple as that.

So Mr McBride did us all a favor in Las Vegas. He showed us his prime evidence, and it’s as close to nothing as something can be. The rest of his claims depend on examples of literal copying that SCO has never shown and implicitly concedes it cannot show. No doubt more swaggering and bluster lie ahead, but we can now be sure that behind it is desperation and greed, not offended virtue.

This column was first published in the UK in Linux User. It is also available in PostScript and PDF formats.

permalink | columns/lu | 2003.09.07-00:00.00

Comments are closed for this story.

Trackbacks are closed for this story.