Michael Dolan Dot Com

Thursday, February 22nd, 2007

Just who contributes to Linux? An in depth analysis of code commits

I had been considering creating something like this myself – I’d almost prefer to build scripts, have them run regularly, store in a database and have a website with the latest stats… but I haven’t found the time (yet).

Back to the analysis, I thought it was very interesting to see IBM’s contributions and leadership in the community. Others.. say… ahem… HP… (hint: look way down the list…) could probably give a little more.  I also find it spectacularly exciting that the largest single group to contribute code to the Linux kernel is still random, individual developers. Apparently they reap enough return from participating in this community to make it worthwhile. I think back to my university days and all the academics studying, analyzing, writing, and fixing Linux code – contributing code back (if it made sense) – that’s what makes Linux so amazingly universal.

I’ve seen other analyses like this in the past and its nice to see Novell now up there at the top. Interesting for 2.6.20 was the rise of Quamranet (the guys behind KVM). Hopefully we’ll see them stay up near the top – KVM could be the “Xen” that actually makes Linux a ‘killer’ virtualization platform. One surprising name on this list was Sony – could they be assisting PS3/Linux?

One group of that’s missing are all the “behind the scenes” developers such as some of the developers in the IBM Linux Technology Center (LTC) that work on fixes and aggregate them together before a single point developer merges a consolidated contribution into the kernel. They are the “unknown soldiers” and I suspect other companies have many of them as well. Read the caveats in the analysis as they have an impact on what each list actually means. It’s easy to take something like this the wrong way.

The Linux developer community is what really keeps Linux steaming along – many vendors and many developers – it’s a train, and it’s very hard to stop a train this big. And this analysis is JUST THE KERNEL. Add in all the libraries, tooling, GUI work, packaging, and other supporting software around the kernel and the community explodes. This is why I laugh when I see “OpenSolaris this or that” from the Sun marketing machine – I took a look and 3 community developers accounted for over 50% of their “community contributions” accepted into the OpenSolaris codebase. That’s for an entire distribution – not just a kernel… I put the community contributions in quotes b/c most of the changes are corrections to comments or minor bug fixes – nothing architecturally significant or evolutionary. OpenSolaris forums are a good Solaris user forums – lots of discussions, users helping users, but very little code. Very similar to Gentoo forums but for Solaris users. Hence even 1.5 yrs later, some changes users have been demanding to make still plague the platform. These issues aren’t new, nor is it just a couple users pushing the idea, but you need developers with architecture (and code commit) control in the community to make anything happen – and generally those developers prefer vendor independence. Anyway, off my soapbox. Someone asked not long ago what my issues were… and that about covers it.

And here’s the full link to the LWN analysis report that I highly recommend reading (all the way through):

http://lwn.net/SubscriberLink/222773/1b982bab71b59bd9/

BTW, I wonder if we’ll see a microsoft.com domain on there soon with the interop labs with Novell… that would be a fun vendor to see on this list next time.