Making separate but critical points about the path of the Linux kernel, the maintainer of the kernel on Monday stressed there is no need to worry about forking and not to expect a move to the GNU General Public License (GPL) version 3.
Speaking at the LinuxWorld Conference & Expo in San Francisco, Andrew Morton, who maintains the Linux 2.6 kernel for the Linux Foundation, also was critical of Sun Microsystems. Sun, he said, has fragmented the non-Windows operating system world with OpenSolaris, which is an open source version of Solaris that rivals Linux.
Acknowledging that fears of forking arise from time to time, Morton he said he did not think it was possible for it to happen, because no one organization contributes enough to the kernel to enable a forking. The level of contributions determines how much a contributor controls.
With a forking, separate lines of the kernel would emerge, which could cause fragmentation and cripple standardization in the platform.
"This is my little attempt to dispel those rumors. I don't see any way in which [forking] could happen," said Morton.
Intel is believed to the top contributor to the kernel and its contribution only amounts to 4 percent, Gordon said.
"That means no single organization has the manpower to take the kernel, run off and fork it," said Morton. Forking would not be economically feasible. The only remote possibility of forking would be if a group of organizations controlling about 40 percent of the kernel joined forces and decided to fork it, Morton said. He stressed he sees no chance of that happening.
What does happen is many organizations, rather than forking the kernel, maintain a private branch to provide additional functionality to their customers, said Morton.
The kernel does endure a massive rate of change, Morton said. "We've been adding or changing 9,000 lines of code per day," during the past five years, he said.
Responding to an audience question about migrating the kernel to GPLv3, which was released on June 29, Morton said he did not see this happening. Linux founder Linus Torvalds is happier with the latest version of GPLv3 than he had been before, but the kernel community is not sufficiently motivated to make the effort to move, said Morton. He noted it has been said that all copyright holders, or contributors to the kernel, would have to sign off on this effort. Linux currently is licensed under GPLv2.
Meanwhile, Morton did not put much stock in Sun's rival OpenSolaris project. "From where I sit, I don't hear much about it. I don't see much evidence of people switching over," or seriously considering OpenSolaris, said Morton.
Sun should have moved off of Solaris and onto Linux, Morton said.
"They've fragmented the non-Windows operating system world and they continue to do so," he said. But he acknowledged he did not see much chance of Sun moving away from Solaris.
Sun could have chosen to move Solaris functionality to Linux, Morton said. There has been interest in moving Solaris ZFS (Zettabyte File System) to Linux but there is an issue with license incompatibility, said Morton. There also has been interest in porting Sun's DTrace technology, he said. Morton said he believes Linux eclipses Solaris in device driver support.
Speaking on the future of the kernel, Morton said there is no overall roadmap. The direction of development is mostly determined by organizations that control kernel developers and they do not have plans, he said. But Morton did predict that virtualization, also known as containerization or resource management, would be a focus in the next one to two years. This technology offers benefits such as machine-partitioning, to partition workloads, as well as server consolidation, Morton said. Technology from the OpenVZ project, for Web server virtualization, is anticipated for the kernel, he added.
An audience member remarked that Novell NetWare has had file permission systems superior to Linux and asked how to get innovations added to Linux.
"I'm wondering what it would take to sort of get people to look at the way things ought to be," the audience member said.
Morton advised that in suggesting improvements to Linux kernel developers, users should present an actual use case of what they want to happen and note that the Linux kernel cannot do it. This frees the hands of developers to implement the user's requirement. The suggested functionality may already be in the kernel, Morton said.
The kernel "is a very large project," Morton said. "It's a very mature project. It's been around for a long time."
Key tasks include fixing bugs and adding support for new device drivers, he said. But there still is a high rate of change in the kernel, Morton said. The distributors of the kernel, such as Red Hat or Debian, are viewed as the main customers, he said.
The kernel undergoes multiple layers of testing before the product reaches an end-user, said Morton.
While embedded development is important for Linux, the kernel team lacks feedback paths from this community, Morton said. "Not many embedded developers really participate in kernel development," said Morton. But he said most of the pressure for features in the kernel is in the enterprise area because the desktop, consumer and embedded areas are addressed fairly well.
Previously, there was a dual path for the kernel, in which a new kernel technology was being developed alongside a stable, productized kernel. But this caused problems, with thousands of patches from the Linux 2.5 kernel being back-ported on top of 2.4 distributions.
Now, consumers are provided with the latest kernel technology and can decide on what to use from a continuous stream of kernel versions.
Morton had little to say about Novell's interoperability agreement with Microsoft over Linux, in which the companies agreed not to sue each other's customers over any intellectual property infringement issues. "I really don't understand [the agreement," he said. Morton said he had not spent time trying to understand what people are upset about in relation to the arrangement and that he had not seen any consequences of the agreement.