![[LWN Logo]](http://old.lwn.net/images/lcorner.png)
![[LWN.net]](http://old.lwn.net/images/main.png)
|
|
|
Bringing you the latest news from the Linux World.
Dedicated to keeping Linux users up-to-date, with concise
news for all interests
 Sections: Main page Security Kernel Distributions On the Desktop Development Commerce Linux in the news Announcements Linux History Letters All in one big page
Other LWN stuff:
Archives/search
Recent features: Here is the permanent site for this page. See also: last week's LWN.
|
Leading items and editorialsA shakeup at MandrakeSoft. At this point, layoffs at Linux companies are not particularly surprising news. But MandrakeSoft, the producer of the Linux-Mandrake distribution, thus far had seemed immune from such problems. It was ![[Mandrake logo]](/2001/0524/mandrake.png)
as if the famous "Mandrake touch" extended to the company's business
dealings as well.
That changed this week, as the word went out that MandrakeSoft had been through a layoff of its own. The initial rumors were scary, but the picture that emerged after the dust settled was more reassuring. What MandrakeSoft is going through now is certainly no fun, but the company may well be better off afterwards. MandrakeSoft has eliminated some 25 positions, leaving the company with about 125 employees. Among the departed is the top management team, including CEO Henri Poole, and much of the e-learning division. MandrakeSoft's developers, including many who work on external free software projects (KDE, the kernel, etc.), are still on the job. There are, says the company, no more layoffs planned. Understanding what's happening here requires just a bit of history. The Linux-Mandrake distribution is relatively new - the first release was put together by Gaël Duval in July of 1998, using Red Hat Linux 5.1 as a starting point. Mr. Duval had no commercial aims at the time - in fact, he took a two-week vacation immediately after the release and was rather surprised by the quantity of mail he had when he got back. It didn't take too long to figure out that Linux-Mandrake had commercial potential, however, and Mr. Duval started MandrakeSoft with a couple of colleagues later that year. MandrakeSoft continued with the idea of making the best, friendliest distribution available, and met with considerable success. The current PC Data numbers for the first quarter of 2001, for example, show Linux-Mandrake at the top of the U.S. market with almost 34% of sales - and that was before the 8.0 release was available. With success came venture capital and a new management team. Henri Poole was brought in as CEO to grow the company and take it toward an initial public offering of stock. But Mr. Poole, it seems, never quite understood the nature of MandrakeSoft's Linux business; instead, he wanted to take the company in the direction of electronic learning and services, and away from Linux. This change in direction did not sit well with the founders and employees of MandrakeSoft. The low point, perhaps, happened with the LinuxWorld announcement of the acquisition of Coursemetric, an electronic survey company which had nothing to do with Linux. The company's board did not go along, and that acquisition was silently canceled later on. In the end, MandrakeSoft and Mr. Poole agreed to part company. As Mr. Duval told us: As soon as it was clear that we wouldn't agree any longer about the strategy, we mutually decided to stop our collaboration. Our core component is the Linux-Mandrake system, an attractive OS targeted both to personal use and server use, with a business around. In other words, MandrakeSoft is back on its original track. Co-founder Jacques Le Marois is back in the CEO's office, and the company is back to its focus on the Linux-Mandrake distribution. Mr. Le Marois will continue running the company for the foreseeable future. The electronic services offerings will remain, but will not be the primary focus of the company. Work will continue toward putting Linux-Mandrake on the desktop, but there will be an increasing emphasis on server deployments as well. MandrakeSoft is also working on "one of the biggest Linux clusters in the world," for a European research agency. There's a set of OEM deals being worked out, including one with HP. The distribution business makes money now (though the company as a whole does not), and a break-even result is possible in the near future. In fact, MandrakeSoft is even planning to restart its IPO process. There is a registration page for those who want further information on the IPO, once that information is available. So the rumors of MandrakeSoft's demise are somewhat premature. Refocusing a company in this way is hard, to say the least, but MandrakeSoft has a good base to build on. By focusing on what it does best, and what has already brought success, MandrakeSoft should be able to position itself as one of the survivors of the current shakeout. They have earned that position. (See also: the full text of our interview with Gaël Duval, this interview with Jacques Le Marois on NewsForge, another interview (in French) with Gaël Duval on LinuxFr.org (English translation available via Babelfish), the MandrakeSoft press release, and this interesting Slashdot comment by an ex-employee of MandrakeSoft). License trouble ahead? Free software depends on the integrity of, and respect for, its licenses. This reliance has worked well so far; most licenses are reasonably clear on their terms, they have mostly been respected, and the occasional problem is generally cleared up fairly quickly. That situation is probably too good to last, however. Even though most licenses attempt to specify exactly what is covered, and under what terms, there are still situations that are not entirely clear. Licensing ambiguity can lead to no end of problems. As more software companies feel the pressure - from the economy, from free software, and from each other - they could be motivated to ignore or attack the rules. It would not be surprising to see an increase in cheating on free software licenses, and on the GPL in particular. There also may be, at some point, a deliberate effort to break, in court, a license like the GPL. Consider, for example, Microsoft's increasing fear of the GPL; that company would certainly not hesitate to send out lawyers to disable the GPL if it thought it would be successful. The following two items look at two situations - one real, one still hypothetical - which show the sort of trouble that the free software community could encounter. The moral of these stories is clear: it is important that we have our act together with licensing. BSD is not free software? IPFilter is the firewalling system used in FreeBSD, OpenBSD, and NetBSD; it provides many of the same features as the Linux "netfilter" package. The IPFilter code is copyrighted by Darren Reed, and has long carried the following license text: Redistribution and use in source and binary forms are permitted provided that this notice is preserved and due credit is given to the original author and the contributors.
A recent update, however, has added the following: Yes, this means that derivitive or modified works are not permitted without the author's prior consent. That means, of course, that IPFilter is not free (or even open source) software. Mr. Reed claims that this is not a license change, that the IPFilter license has never allowed the modification of the code. This, of course, has come as a surprise to some users, who had assumed that an important part of BSD would be distributed under the BSD license. It is also not surprising that, in a free software climate, people might interpret "use" to include modification. This is not a theoretical issue. OpenBSD distributes a modified version of IPFilter; the modifications, they say, are needed to make IPFilter work properly with their system. OpenBSD leader Theo de Raadt has told us he does not believe the "reinterpretation" of the license is valid. Mr. de Raadt is not known for backing down from confrontations, so an interesting legal situation could develop here. Mr. Reed is unconcerned about potential problems with his license. What if he gets hit by a bus and no longer distributes updates to IPFilter? "I won't care, I'll finally get to RIP." He also refused to answer questions from LWN ("I really don't like Linux"). There is only so much room for criticism of Darren Reed, however. The author of the code gets to pick the license. The real problem is that code with a non-free license was incorporated into the core of a free operating system. Carelessness with licenses invites trouble. The boundaries of the GPL. It is tempting to say that the above situation would never happen with the Linux kernel. After all, the kernel is licensed under the GPL, which would have prohibited the incorporation of code carrying the IPFilter license in the first place. But the Linux kernel, too, carries a potential time bomb with its licensing. It is a standard feature of the GPL that the code it covers can not be linked with code that does not carry the GPL's protection. This is the so-called "viral" feature of the license. But it is always an interesting question to determine what constitutes "linking," and what is "mere aggregation." Linking code into the kernel code itself clearly requires the linked code to be licensed under the GPL. On the other hand, a program that runs in user space is not considered linked, despite the fact that it relies on the kernel for operating system services. But what about loadable kernel modules? They are linked directly into the kernel image (at run time), and run in kernel space. Need they be licensed under the GPL? Linus Torvalds answered this question many years ago through an "interpretation" of the license; in essense, as long as loadable modules limit themselves to the "published" interface, they need not be free software. So proprietary device drivers are possible. It is also possible to add a great deal of other functionality into the kernel using the loadable module mechanism - this is a reasonably large loophole. It is also a problematic one, for a couple of reasons. The first is that the published interface that may be used by proprietary modules is not particularly well documented. Presumably it means that a proprietary module may only access data structures and functions which have been explicitly exported to modules. Linus also once stated that modules are not allowed to add system calls. But the fact remains that a great deal of functionality could be added to the kernel via a proprietary module, and it is not clear just where it crosses the licensing line. But the real problem is that Linus does not own the copyright for the entire kernel. Many major contributors have retained their own copyright on the code they have added, and many of them are opposed to proprietary modules. That leads to a couple of troublesome scenarios:
Free software licenses have served the community well for almost two decades, and will certainly continue to do so for a long time. But a certain degree of attention and vigilance is clearly called for, or licensing will lead to ugly messes that somebody has to clean up. Inside this week's Linux Weekly News:
This Week's LWN was brought to you by:
|
May 24, 2001
|
Copyright © 2001
Eklektix, Inc., all rights reserved
Next: Security