LinuxCon: Roundtable – The Linux Kernel: Straight from the Source (liveblog)
September 21, 2009 1 Comment
I’ll be liveblogging from LinuxCon here in Portland. I have not been posting as many traditional blog posts recently, something I’d like to remedy after LinuxCon. Stay tuned.
Panel: Bottomley (moderator), Jonathan Corbet, Greg KH, Linus Torvalds, Arjan van de Ven (detained in Holland – NP), Ted T’so and Chris Wright
Opinion on what’s the most innovative recent kernel feature:
Chris – virtualization
Jon – ftrace and performance counter framework
Greg – USB 3.0 (best thing to come out of staging: a working laptop for Linus)
Ted – expanded on Jon’s performance counters answer and then added kernel mode switching
Linus: his job has gotten much easier in the last 3 months. He really likes this… (added: “Xen will have a difficult time merging their tree into mainline as-is”)
Linus: Over the last 18 years, what has been the most inspiring or motivational aspect of the Linux kernel?
* started out being all about the technology, then become more about community and even the “fame”; these days it’s “all about the Linux kernel community”. “I really enjoy arguing”
* “The Linux kernel is a life-long calling for me”
Bottomley added: “Interestingly, the average age of kernel maintainers is continuing to rise. How do we guarantee kernel development continuity long term.
* Linus: There continues to be young people getting involved.
* Greg notes that in many cases, they don’t even know the age of maintainers
* Jon: “I don’t think we’ll lack for talent”
* Ted: roughly 50-60% of people going to the kernel summit are first timers. At the very top, however, people have remained fairly constant over time
* Chris: as we add more subsystem maintainers, people are getting more niche. It takes a fairly motivated person to get involved as a maintainer.
As the rate of kernel contributions increase and as the kernel becomes higher profile, is it getting more difficult to keep out malicious code?
* Greg: We do currently track regressions
* Linus: Our issues have never been intentionally malicious, they’ve been unintentional bugs. The only worry he’s had about malicious people, he addressed in git by cryptographically signing the public repos. This was a result of someone breaking into the bitkeeper repo years ago and being caught.
Follow up by Bottomly: as we add more code more quickly, performance has been going down 1-2% a release, with a 12% degradation in the recent past. How do we address this?
* Linus: “I’d like to tell you I had a plan”. Admits there is some bloat but says part of the issue is possibly unavoidable.
question about the state of the current sound subsystem
* Linus: “The sound subsystem isn’t as bad as some people make it out to be – don’t listen to the crazies on slashdot”.
* No sound maintainers on the panel. Dave Phillips would be a good person to ask.
* Jon: “Sound is a mess in a lot of ways. A lot of flux and professionals don’t like Pulseaudio. latency is an issue.” “I do think things are starting to get better”
* Greg: A lot of new mixing boards and actually now running embedded Linux
What would you like to see in the Linux kernel, but feel may not be feasible?
* Linus: “We’ve never hit a problem that we felt was impossible to implement and generally useful”
* Ted: “Speaking with the Microsoft NTFS team at Redmond, they have actually come up with a system remarkably close to the Linux kernel development model”
Note: 2 out of 5 of the panelists read “almost every message on LKML”. Linus was not one of them (Greg KH and Jon)
Will next year be the year of the Linux desktop?
* Ted: “Next year will be the year of the Linux desktop because ‘next year’ is always the year of the Linux desktop.”