IBM fingered over early Linux mistakes

Agenda first, kernel second

Linux FCS IBM, the first and biggest IT vendor to champion Linux, has been gently rebuked for initially tolerating the community to further its own interests.

And IBM was forced to concede.

IBM was identified during a Linux Foundation conference debate on community participation as emblematic of big companies who've made the mistake of joining the efforts around Linux to further their own goals first, and to help development of the kernel second.

James Bottomley, Novell distinguished engineer, kernel maintainer, and chairman of the Foundation's technical advisory board, said IBM initially approached Linux with the belief it had key technologies that should be in Linux, and pushed them upstream into the kernel.

These technologies, he said, first benefited IBM and only secondarily benefited the kernel.

"They broke down the technology into elements that would benefit IBM," Bottomley said.

He did not identify the technologies involved, but IBM engineers and expertise have been poured into making Linux work on its Power architecture and Intel-based servers. Back before the wider world understood Linux, IBM was there with its own Technology Center (LTC) in 1999, manned by 185 IBMers in six countries. IBM's server and global services business has since become one of the biggest beneficiaries of making the kernel - via Red Hat and SuSE - work well on its hardware.

How did IBM respond to this characterization? It agreed.

The vice president for open systems development in IBM's systems and technology group Dan Frye said: "Absolutely - we learned a lot from that lesson." What IBM learned exactly was how to join and participate more effectively in the community, he said.

Clearly there was a lot of this kind of thinking going around in the late 1999s and early 2000s. IBM created the Eclipse open-source tools project in late 2001 with a major code-dump from its Visual Age for Java tools, and with backing by its engineers.

The goal was to stop the pointless and costly task of re-building basic development-tools frameworks and tap a broader market of plug-in providers. Also, to outmaneuver tools and Java competitors.

The project worked better than expected on all fronts. But eight years later IBM remains the single largest participant in the resulting Eclipse Foundation and its projects.

The whole subject came up during a panel debate on how to measure community contributions to Linux during the Linux Foundation's Collaboration Summit in San Francisco. Along with Frye and Bottomley, participating were Red Hat developer-community manager and Fedora Project board member Karsten Wade, and Ubuntu community manager Jono Bacon.

The debate had turned to how to encourage participation and mistakes commonly made. According to Bottomley, there are open-source projects that claim to be open but only involve a "tiny bit of community" - and nobody knows who they are.

The biggest success of Linux was drawing on more than just a core set of contributors, he said. "It's almost the lifeblood of the OS," he explained.

Frye claimed there exist other community projects that let you only scratch their particular itch - or solve a specific problem that only the participants are interested in, and that have little broader benefit.

Again, there were no names.

Wade said companies had to learn that they couldn't apply the closed-source model of software development to open source by hiring people and then telling them to code. He called this stacking a project by hiring all the people to make sure the programming goals are met.

"[Companies] need to focus on not stacking-up a project by hiring all the people to make it happen," Wade said. Bottomley called this an "industrial process" of paying people to achieve a goal that's different from the community approach of working collaboratively. ®