SAN FRANCISCO — It may be thankless, boring work, but you have to get more involved in building the Linux Kernel.
That was the underlying message to Sun Microsystems and Microsoft by Linux Kernel co-maintainer Andrew Morton during a keynote address at the Linuxworld conference here.
His principal theme, however, was one of participation in the process of building the Linux kernel, something that he referred to as a moral obligation of Linux users.
“The kernel is a very dull project and that’s the way we want it to be,” Morton said.
Morton explained that it is the Linux kernel that defines what a Linux system is. He told the assembled Linuxworld audience that you can take all the hardware of every last atom and you’ve still got Linux. Replace all the software above the kernel and it is still a Linux system.
The kernel is a rapidly developing project, though Morton characterized Linux kernel development as largely a maintenance exercise fixing bugs, cleaning up the code base and adding drivers.
“It remains the case that feature pressure on the kernel is very high,” Morton said. “I don’t know when it will slow down.”
The feature pressure is for enterprise features on server machines with a lot of work done on desktop workloads, consumer and embedded devices.
Morton noted that the some 9,000 lines of code are added or changed per day. In between the 2.6.21 and the 2.6.22 kernel, some 6500 code commits were made, which he characterized as typical.
“There is always a trade off between how much you change and how many bugs you introduce,” Morton said. “Kernel developers have made the decision for the high change rate. I think this is the correct decision. It means we have other groups downstream from us that do additional release engineering to make the kernel stable for an end user.”
Though the kernel with its high rate of change may not always be exactly what every stakeholder is looking for, Morton was adamant that the kernel itself would not fork.
“I don’t see any way it could happen,” Morton said. “Why it will never fork is because none of the organizations that contribute write more than a tiny fraction of the code that goes into it.”
By Morton’s count Intel is the largest code contributor at 4 percent of the Linux kernel. That means no single organization would have the manpower to create a fork. The only possible scenario would be if a group of organizations, representing 30 to 40 percent of code, joined up to fork the project. Morton sees no prospect of that happening.
“If people are unhappy with the direction, rather than forking they can come and tell us and we can make adjustments rather than making a great split.”
It is people telling kernel developers what is wrong that is critical to the development of the Linux kernel. Morton said testing the kernel is a great way for people to contribute and help out the entire Linux world. He said that users should test the kernel out of the goodness of their hearts and as an in kind payment for getting Linux.
“We’d be absolutely screwed without those crazy people that download and test the kernel.org kernel,” Morton said. “It is Linux’s sole competitive advantage and is the reason why we are where we are today.”
As to how the kernel itself develops new features, Morton admitted that the core kernel development team has no plan, and that it’s not a bad thing. “It means the kernel development is reactive and proceeds in the direction of greatest need.”