dcsimg

Who’s to Blame for the Linux Kernel?

Ever wonder who contributes the most to the Linux kernel? Of course you have. Here's a hint: It's not Canonical, certainly not Microsoft, and you might be surprised which companies crack the top 20 and where.

Finger-pointing time! Let’s see who’s responsible for kernel development in the last year. Once again, the Linux foundation has released its report on who wrote Linux. As always, it has some interesting insight into who did what when it comes to kernel development, and the direction of the kernel. Unsurprisingly, embedded/mobile is becoming a major factor in kernel development.

The Linux Foundation publishes an annual Linux report that shows (approximately) who has written and contributed to the Linux kernel. The report is put together by LWN’s Jon Corbet (also a kernel contributor) and kernel developer Greg Kroah-Hartman, with additional contributions from the Linux Foundation’s Amanda McPherson.

The Top 5

Everybody wants to know, who’s at the top of the list. Consistently at the top is “none,” which is to say that nearly 20% of the kernel development is done by people who aren’t affiliated with a company — at least as far as their kernel contributions go. Yes, Virginia, independent kernel contributions still exist.

The report provides two lists — contributions since 2.6.12, when Git logs became available, and since the last report (2.6.30). Red Hat tops both lists, with 12.4% of kernel changes since 2.6.12, and 12.0% since 2.6.30. A tiny decline, but remember that the number of developers participating in each release cycle grows by about 10%. Meaning that the proverbial pond keeps getting bigger, and the Red Hat fish isn’t getting much smaller in comparison.

The red fish keeps growing, but the green fish isn’t keeping up quite as well. Novell had 7.0% of kernel contributions since 2.6.12, but only 5.0% since 2.6.30. It’s dropped from second to third in kernel contributions, after Intel, which had 7.8% of kernel contributions since 2.6.30. Some of that may be that more X.org is being moved into the kernel, and a lot of X.org development is being done by Intel, and Intel is also doing more with its work on MeeGo.

Intel comes in second on most recent contributions, bumping Novell to its third place spot. IBM is also displaced by Intel, landing at fourth (Intel’s old slot). Who’s in fifth (sorry Abbot, Costello)? Nokia. Yep, Nokia — who were behind SGI, Parallels, and Fujitsu in 2009.

If you’re looking for individuals, the top five since 2.6.30 are Paul Mundt, Johannes Berg, Peter Zijlstra, Bartlomiej Zolnierkiewicz, Greg Kroah-Hartman. Mundt explains Renesas’ place in the list — he’s working for them, after a stint at the CE Linux Forum (CELF). Berg is on Intel’s payroll, working on wireless, Zijlstra works for Red Hat, and Zolnierkiewicz is a student at Warsaw University of Technology. Kroah-Hartman, of course, is at Novell.

Linus Torvalds doesn’t make the list not because he’s not doing anything, but because the list doesn’t measure what Torvalds does very well. That is to say, Torvalds spends much of his time merging commits from others and not so much writing his own code. Still quite important, but not as easily measured.

I beat Oracle up pretty heavily lately because of their antagonism towards Google and open source Java, as well as their mishandling of OpenSolaris, OpenOffice.org, and virtually all of the properties they got from Sun. Nothing that’s related to open source has gotten better since Oracle took it over. Still, the company turns in a respectable — if somewhat reduced — showing in kernel development. Oracle clocks in with 1.9% of kernel changes since 2.6.30, and 2.3% since 2.6.12.

Then there’s Canonical. Or rather, there Canonical isn’t. Once again, the most popular Linux desktop vendor and would-be enterprise Linux player doesn’t rank highly enough in kernel development to show up — even in the past year. I might get flamed for mentioning this, but I do think it’s worth pointing out. Yes, Canonical makes valuable contributions to Linux in other areas — even if the seem ashamed or reluctant to mention that Ubuntu is Linux underneath. Does Canonical need to contribute to the kernel to be successful? Apparently not. Should Canonical be contributing more given its standing and dependency on the Linux kernel? I believe so.

Embedded

Nokia’s placement on the list shows that much more development is being driven by mobile and embedded Linux. In the past, server Linux was the big money behind the kernel. Still is, but it’s making room for embedded Linux.

Nokia has jumped up in the standings and has doubled its percentage of contribution. Wolfson Microelectronics and Renesas Technology appear in the top 20 for the first time. Both companies are working with embedded Linux. Texas Instruments also makes the list — Linux on a calculator, anyone?

Broadcom and Atheros also make the top 20 since 2.6.30 — which is good, we might see fewer and fewer chipsets that aren’t supported in Linux.

What’s disappointing is that Google isn’t higher in the ranks here. Actually — Google has dropped off the top 20 altogether since 2.6.30. The search giant had less than a percent (0.8%) of kernel changes since 2.6.12, and only 0.7% since 2.6.30. Google is behind Pengutronix, for goodness sakes. Have you heard of Pengutronix? Nope, me either. For a company that is arguably using more Linux than anybody — pushing two Linux-based OSes and likely to have more Linux servers in use than any other entity — Google’s kernel contributions are actually quite paltry.

Summary

2011 should be interesting. If Google finally merges Android’s changes into the mainline kernel, that should bump Google up in the standings. I suspect, and hope, SUSE/Novell will move past Intel in 2011, now that its future is a bit more clear. As MeeGo continues to gather steam, I suspect Nokia will also show up a bit higher in the standings.

In all, Linux kernel development is as healthy as ever. I’d be curious to see a similar report for other major system utilities and such (GCC, the GNU utilities, X.org, Apache Web server). The kernel is very important, but just a part of the overall ecosystem. There’s plenty of userspace goodies that companies should get credit for as well.

Make sure to check out the full report PDF too. It makes for good reading, and it’s short and well-written.

Fatal error: Call to undefined function aa_author_bios() in /opt/apache/dms/b2b/linux-mag.com/site/www/htdocs/wp-content/themes/linuxmag/single.php on line 62