When asked what is the one thing Linus is proud of that Linux has achieved, it is that it has enabled people to achieve all kinds of things (e.g. OLPC) that Linus would never have done himself.
John Corbet said that Linux is now driving in the dark, it doesn’t have anyone to follow anymore. So let’s ask the guy at the steering wheel where we’re going. Linus says not to worry about the kernel. Linus never had a real vision, the direction has always come from outside (because it was already perfect for Linus in ’91).
The kernel people like arguing, and sometimes personalities really clash. This worries Linus. Also the broken patent situation worries Linus. But technically Linus has no worries because there’s a very good community to solve them. Occasionally he worries about hardware companies because they’re hard to work with. The problem is that the traditional way of working of hardware companies (develop in isolation for 1-5 years, then release software and keep it static for 10 years) doesn’t match the reality of Linux. However, hardware companies that change their approach and really track the kernel are generally happy with that. The kernel model of global changes is going to stay.
Specialized systems (big iron or small embedded) doesn’t see the same amount of testing as the usual PC and server environments. This is potentially dangerous for those exotic systems, but Linus doesn’t see a solution. He’d like to have the opportunity to do testing on a 4096 core machine 🙂
Will there be a new disruptive OS in the next 10 years, or is Linux it. Linus thinks that OSes don’t need that much innovation. Sometimes the old ways are just the correct way. What OSes should do is to work well, not to innovate. All major OSes today are based on ideas from 50 years ago. In addition, trying to compete with the huge brain tank is very hard. But Linus is a bit biased.
Is there a procedure to remove stuff from the kernel that nobody is using anymore? But there’s a very high threshold to remove things, because it occasionally happens that somebody has access to a piece of hardware and finds that the kernel almost works, then steps up as a maintainer. The idea is that the cost of maintaining such dead code with known bugs is not high. Sometimes there are real overheads that e.g. the security people have added and has to keep on being supported.
Android: is not very different from the RedHat situation ten years ago: they maintained their own fork. A year ago, the kernel community decided that Android has proven that it works, and because of applications the APIs will need to stay, so the Google changes are being accepted. Since distros only recently started upstreaming first, we can expect it may go the same way with Google.
Performance vs. power: not really a conflict, it mainly goes in the same direction. Power has extra issues, but not much conflicting. People have become very conscious about power, both the kernel people and the hardware manufacturer. Also power measurements are becoming better, which makes debugging power issues easier.
So what does the future bring? “We have no plan, and everyone is happy”.