Why Open Source Will Rule Scientific Computer
Dr. Will Schroeder stopped by PNNL today to talk about his company Kitware, their process and business strategy involving Open Source software, and how the world is evolving toward OS. He is giving a similar presentation at Oscon in Portland this week.
I got a chance to talk to Will beforehand. He brought up CMake, which his company created and maintains, which gave me a moment of "hey, I've used that!". He was quick to mention the scarcity-centered mindset of companies that make closed code and how we have such infinite problems to solve, we need to remember there's plenty for all of us.
Sells Expensive Services, Gives Away Free Software
Will went into detail sharing Kitware's chronicles as a professional services company in high performance computing, medical imaging, computer vision, and other fields. He mentioned numerous projects, including the Visual Tookit (VTK), along with appealing images produced by their software. There was an impressive set of collaborating institutions and organizations mentioned, including Harvard, IBM, and Sandia National Lab.
Fundamentally, Kitware provides Open Source collaboration frameworks and tools knowing customers will need them supported and customized. They then sell the service of developing and maintaining those customizations. It's a basic business strategy for Open Source, but one Will emphasized works for Kitware and lowers cost of doing business for everyone. I'm sure Carl Cadwell's view of having a "volume" and "margin" product plays into the strategy here as well.
The company also focuses on a very rigorous, test-centric Agile process. Their team works in 2-4 week sprints for building code, but they test it nightly using unit testing across different platforms and sites, each collaborating entity seeming to have their own results centralized and publicly displayed. They always focus on having a dashboard to display these results to supplement more mundance collaboration tools like mailing lists and Wikis.
The best moment was when he brought up the Ohloh statistics for Kitware's codebases. According to Ohloh's estimator, the company has given away over 138 millions dollars worth of code. Will, and his company, has really invested in Open Source to create what appears to be a thriving company.
Problems are getting so big now...we have to work with others to get things done
Will then transition into the grander topic of Open Source and its relationship with industry. He pointed out that Open Source enables communities to maintain software instead of companies. This means more eyes and minds working on the problem, with lower costs for each participant. As one audience member pointed out, this isn't true for every SourceForge project released by three guys (well, he said "physicists", but who wants to go around pointing fingers about who makes crappy code), but "communities instead of companies" is a classic reason for why Open Source has a good reputation for quality amongst its supporters.
Seeming to be a business-focused individual, Will did point out that copyleft licenses (such as the GPL) do tend to scare off mainstream business partners. He prefers permissive (MIT, Apache) licenses since it enables a "no string attached" outlook for business. "If a give has a string on it, it's not really a gift" he pointed out. This seems to be the norm for for-profit companies hooking into Open Source. They want to sell services, not litigate over licenses.
It was great to meet and listen to Will and his take on Open Source. He presented excellent arguments for Open Source, detailing how it's cheaper, higher quality, and possible to monetize. He also discussed how the general evolution in business and particularly research is toward Open Source. Just like how scientists work in a larger bazaar of ideas, shared freely so discovery can build on discovery, scientific software needs to follow the same ideals.