As you will recall, SCO's expert, Dr. Thomas Cargill, has opined that "Linux is a substantial copy of UNIX System V Release 4 ("SVr4") because it appropriated the essential structure of UNIX by incorporating (1) many of the "system calls" in SVr4; (2) the SVr4 file system; (3) the ELF format; and (4) the Streams communication module. (Id. at 3-4.)." On that basis, SCO is claiming copyright infringement.
Groklaw has already presented information answering their claims regarding ELF, STREAMS and the Linux file system, so I started thinking about system calls.
I remembered the Linux Kernel Personality, or LKP. Surely, I thought, this isn't news to SCO about system calls. As Caldera, they wrote the Linux Kernel Personality, which is a way to run Linux applications and binaries on a Unix kernel, so they surely had to know what was inside the Linux kernel. After all, they had both kernels and had them side by side, so they had to notice any copyright infringement on the spot. That goes back to 2001, as I'll be showing you, and Caldera, according to a news report from the time, was testing the Linux 2.4.1 kernel, which SCO now claims infringes.
I also found a Caldera PowerPoint presentation on LKP, with a section on system calls, which I thought had disappeared from the Internet, and some SCO documentation on the same thing, and I took some screenshots for you of the latter, since we have found that when Groklaw writes about something on SCO's servers, even if only in the Wayback archive, it has a tendency to die and go to the Great 404 in the Sky shortly thereafter.
And if you want to know why I bother to keep digging, it's because unlike the media, I know this is the calm before the next storm, and SCO isn't done fighting. Jungle drums tell me that SCO will be trying to appeal Judge Brooke Wells' Order, filing something tomorrow. Jungle drums are not always right, of course, but I did expect them to try to appeal so I believe it, because they kind of have to, I think, considering how devastating the most recent order was.
Well. They could just admit they have no case, but what are the chances of something as refreshing as that? So, onward we slog.
Way back in 2003, I wrote about LKP, and I directed readers to a PowerPoint Caldera did on the subject. Of course the link I used then no longer works, having joined the disappeared, along with so many of its brothers and sisters. There seems to be no surer way to kill a page on SCO's servers than to mention it on Groklaw.
However, they are not always so careful, and sometimes they leave evidence carelessly scattered about. That is the case here. You can still obtain that PowerPoint presentation, "Linux Kernel Personality Internals," I discovered here [note if you click, it begins the download, so don't click if you don't want that to happen]. And if you look at slides 4-9, you'll find all the information about system calls. The rest is interesting too, because they describe just how hard it was to get two different personalities, as they describe it, to work together. And it mentions ELF and file systems too. The presentation concludes with a slide that says that Open UNIX can run Linux applications and binaries without recompilation:
All it takes is around 40,000 lines of code to build:
- A kernel module for handling system calls
Three totally new Open UNIX filesystems
A new system Daemon
And the infrastructure necessary to install all of these, and the whole of Linux, onto Open UNIX
This, of course, raises the question of whether Open UNIX was released in totality under the GPL, or the simpler question of whether it violated the GPL, which Peter Galli wrote about back in September of 2003. I'll leave that to the experts. Just putting up the raw materials.
[ Update: A reader sent in the following urls to more information on LKP from SCO documentation, http://
uw713doc.sco.com/en/LX_uw/LKP_works.html and from those pages I found this interesting tidbit:
Prior to Release 8.0.1, the /linux/proc filesystem was supported by the lxprocfs kernel module. With Release 8.0.1, the lxprocfs functionality has been merged into the procfs module, and the lxprocfs module has been removed. The /linux/proc filesystem is now mounted as a procfs filesystem with a new mount option (-olinux).]
Here's an article from 2001, by Steven J. Vaughan-Nichols, talking about Ransom Love, then CEO of Caldera, and his plans to merge UnixWare and Linux:
By mid-March, Caldera expects to disclose new partner programs to UnixWare resellers. "We want to make sure that [SCO's] partners know where they fit so that when the transaction is done, they'll know exactly where they stand," says Love.
Caldera executives say partners should expect a new product-branding strategy. Specifically, Caldera's platforms will be branded by functionality (database server, Web server, etc.) instead of by operating system. The partner push will involve cross-selling and cross-development between the UnixWare and Linux communities.
For instance, Caldera will make UnixWare binary-compatible with Linux, allowing UnixWare customers to run Linux applications, Love says.
On the flip side, Caldera Linux will gain UnixWare's best enterprise and database management features. These include large file system support, asynchronous input/output (I/O), the UnixWare API, extended developer kit and multipath I/0. Along the way, Caldera plans to gradually phase out OpenServer, SCO's older Unix product.
Love believes UnixWare extensions to Linux will help major database players, like Oracle, to increase application scalability on Linux.
Caldera also is testing Linux 2.4.1, but the company isn't rushing to push the new kernel out the door.
Zounds, I hear you say. Does this mean SCO is now suing over things it itself wanted in Linux?
I think it's fair to say, then, at a minimum that Caldera knew the innards of the Linux kernel and system calls since at least 2001. They had the Unix kernel, and they had the Linux kernel, and they were looking deeply into the question of how to make the two operating systems merge, with each benefiting from the other, as they saw it. Whatever was the same or similar, methods, concepts, the works, they had before their eyes in 2001, and they were looking. The time to file a copyright infringement claim, then, was back then, if they had objections. The ending to the article is interesting too, from an economic-damages point of view:
Critics note that UnixWare never gained critical mass under its former owners. The product started out as a joint effort between AT&T and Novell. Novell went on to acquire AT&T's Unix business, and tried to merge UnixWare with NetWare. But the project turned into a debacle, and Novell ultimately sold UnixWare to SCO for pennies on the dollar.
At this point, UnixWare certainly could use a little Love.
Interesting, huh? But getting back to systems calls, here are the screenshots I took regarding Linux system calls, and you'll note the date of the SCO article is January of 2002:
As you can see, the page states proudly that the LKP supports "the vast majority of Linux system calls". You can't support something without knowing what is in there. It's just that simple. Here's the list of the few system calls that were not supported on the date of this document, which means someone made a list and checked it twice, so to speak, so it is evidence that they knew precisely what system calls Linux had and what they were like.
Note that while not all the system calls were supported at the time of this document's creation in 2002, the bottom of the page tells us that they would be in the future, so they still wanted it to happen in 2002:
So, what do you conclude about all this talk about copyright infringement? That it's silly? Me too.
Now, here's the serious part. We know SCO reads Groklaw. Heaven only knows they bash it enough when talking to reporters. And more significantly, we know their lawyers read Groklaw, because they have told reporters so at least twice. Here's the most recent on Law.com. (By the way, one correction on that article. People do know where I've worked. They know I worked at OSRM. It was in the papers. So they know I'm a "real person". The media likes dramatics, but in real life, anyone could reach me when I was at OSRM, just like any other person.) And here's the article in the Washington Post by Jonathan Krim in June of 2004, in which he reported they said they use Groklaw as a resource.
So if they persist with these claims, in the face of this evidence on Groklaw, what might it indicate? Is there not a point where a claim becomes actually frivolous? I'll let you draw your own conclusions.
Update: An anonymous reader left the following humorous summing up of SCO's claim:
If it beeps like a car...
"Rolls Royce is a substantial copy of Chevy Nova because it appropriated
the essential structure of Cars by incorporating (1) many of the
"Wheels" in Car; (2) the Steering Wheel; (3) the Wind Shield; and (4)
the Gas Tank module. (Id. at 3-4.)."