Mike Olson was an open source pioneer. But he's not an open source zealot.
Olson helped build the open source Berkeley DB database in the early 90s -- before the Linux boom -- and as the CEO of Sleepycat Software, he turned the database into a successful business using something very similar to the GPL, the free software license that was so essential to the rise of Linux. The GPL -- or GNU General Public license -- said that if someone modified free software and distributed the code with a larger product, they would have to contribute their work back to the community.
But in 2009, Olson founded Cloudera -- the first outfit to commercialize Hadoop, the open source data-crunching platform based on Google's software infrastructure -- and he dropped the GPL in favor of the Apache license, a license that doesn't require contributions back to the community. You might think that such licenses would stunt the growth of the open source world. But Olson believes the opposite is true.
The open source zealots don't agree. But Olson is by no means the only one backing the Apache license. According to some statistics, the world's open source projects are gradually moving away from restrictive licenses such as the GPL and towards more permissive licenses such as the Apache, and many open source watchers -- including longtime pundit Matt Asay and Steven O'Grady, an analyst with developer-minded research firm RedMonk -- agree that this shift will ultimately bring open source to a much wider audience.
"The business-side of open source has moved towards the Apache license model...and Mike Olson was out in front of everyone on that," says Asay, who for many years downplayed the importance of the Apache license. "He was the guy, behind the scenes, constantly calling B.S. on all of us who believed in the GPL."
The main forces behind this movement are the giants of the web, including Facebook and Twitter, which have a very different attitude towards open source than traditional software companies. But as these big names release more and more projects under permissive licenses, they've spawned a new breed of open source vendor, such as Cloudera, that's potentially much more attractive to the world's businesses because it avoids the GPL and other restrictive licenses.
Rightly or wrongly, many businesses are afraid of the GPL, worried it will force them to give up their proprietary code. But with the Apache license, that fear goes away. For Olson and others, this not only encourages the adoption of open source tools, it's a better way for companies like Cloudera to actually make money from free software.
Apache Rising
Restrictive licenses such as the GPL are still the most widely used -- by far -- according to Black Duck Software, a company that tracks the use of free software licenses. But now that the free software movement has matured -- and the web has changed the economics of the movement -- Black Duck's stats indicate that the GPL's influence is waning in a favor of more permissive licenses such as the Apache.
According to the company's numbers, the percentage of projects using the GPL dropped from 70 percent in June 2008 to about 57 percent today, while the Apache and the MIT -- another permissive license -- have risen to 5 and 11 percent respectively.
Brett Smith -- the license compliance engineer at the Free Software Foundation, the not-for-profit that spawned the GPL -- argues that stats like Black Duck's are misleading. "It's been hard to really figure out what those numbers really mean because they come from data that's not fully published," he says. "It's hard to tell what really going on." But others -- such a Redmonk's Steven O'Grady and Paula Hunter, executive director of the OuterCurve Foundation, a Microsoft-backed open source advocate -- say they're seeing the same trend that Black Duck sees. (Black Duck declined to be interviewed for this story).
>"\[Web companies\] have a very different attitude towards open source than we've seen in the past. They don't value code in the same way."
What's clear is that over the last few years, many of the highest profile open source projects have chosen the Apache license, including "cloud computing" platforms such as Hadoop, OpenStack, Cassandra, and CloudFoundry. Node.js, another of-the-moment cloud platform, uses the MIT License. And even the big-name mobile platforms have joined the crowd. Google's Android mobile OS used the Apache license, and just this week, HP announced its schedule for open sourcing Palm's webOS platform under the Apache.
It's no coincidence that many of these projects grew out of the big web companies. "They have a very different attitude towards open source than we've seen in the past," says Steven O'Grady. "They don't value code in the same way. These companies are taking code that would have been proprietary five or six years ago -- that would have been seen as differentiating code -- and just releasing it. They don't necessarily want or need the protections of a restrictive license."
Companies such as Facebook and Yahoo -- the companies that bootstrapped Hadoop -- aren't in the business of selling software. But that's only part of the explanation. They built their operations using open source software they pulled from the community, so they're happy to give back to others without requiring something in return. But at the same time, they realize that others feel the same way. They know a return will come anyway.
O'Grady points to Twitter as another prime example, with projects such as the open source database FlockDB and the open source web developer toolkit Bootstrap. Then there's Rackspace with OpenStack, a platform for serving up virtual computing resources a la Amazon Web Services.
But the trend doesn't stop there. Myriad businesses have sprung up around these open source projects, hoping to help the rest of the world adopt the software -- and make some money in the process. Facebook's open source Cassandra database spawned the Texas-based DataStax. Hadoop gave rise to not only to Cloudera but a Yahoo spinoff dubbed Hortonworks. Rackspace formed its own services division around OpenStack. And Joyent, the steward of Node.js, is selling software and services to companies using the open source development platform. Mike Olson is by no means alone.
The Poison and the Antidote
When Olson was at Sleepycat, Berkeley DB carried a "strong copyleft" license based on the GPL. You could use Berkeley DB for free, but if you did, you might have to pay the cost with your own code. For many companies, it was an unsettling proposition. Though they wanted Berkeley DB, they didn't want to let go of software they'd spent years developing. But Sleepycat offered a loophole. If you paid Olson and company some money, they would give you a separate license that let you keep your own code. It's a technique known as dual-licensing.
>"The GPL was kind of the poison, and we would sell you the antidote."
"The GPL was kind of the poison, and we would sell you the antidote. If you preferred not to infect your source code with the GPL, you could buy a different license," Olson says. "This was reasonably successful for us, but we were never going to be a $100 million-a-year company. Our commercial transaction with our customers was founded on a threat: 'Either you give me some money or I’m going to infect your intellectual property.' That’s not a real good place to start a business conversation."
In using the Apache license, Cloudera completely changes the dynamic. In essence, you can use the free code however you like -- without contributing any of your own code back to the community. Cloudera makes its money by selling support and additional proprietary software that works in tandem with Hadoop. It's a strategy bridges the gap between free software and un-free software.
"Open source is a really important part of what we do. About half of our engineering spend goes to the open source [Hadoop] project," Olson says. "But it’s important to differentiate ourselves from the rest of the market, to have a reason for customers to come uniquely to us."
Some people call Cloudera an "open core" company. The core of the project is open source, and the for-pay software that Cloudera offers around it is not. But the name doesn't come up as often as it once did. "Open core" developed a negative connotation among the community because it implied a lack of openness.
Mike Olson doesn't care what you call it. "Whatever you call it," he says, "it works."
Open Source Pragmatism
The still private Cloudera says very little about it finances, but its list of customers includes such as names as Groupon, Rackspace, and Samsung, and according to Olson, it has made inroads beyond the tech world, on Wall Street and with biomedical outfits. A big part of the company's success, he says, is the Apache license. He understands why people like Free Software Foundation founder Richard Stallman back the GPL -- and, yes, the GPL was very good to Olson for many years. But these are new times.
"I don't believe that a political or religious conviction is the sound basis for a business," he says. "If you talk to some of the early free software folks, it was all about rights and responsibilities -- and I understand why Richard focused on that stuff -- but if you want to build a business, you have to focus on customers and markets and commercial opportunity. The GPL gives you some ways to monetize what you do. But there are other collaborative licenses that give you more."
>"I don't believe that a political or religious conviction is the sound basis for a business"
The advantage is that the Apache is less threatening to potential customers. "If you want software to be used, you want an Apache license," says Mark Radcliffe, a partner with international law firm DLA Piper who specializes in open source software. "There's very little legal complexity for people to deal with." And once you have the users, Olson, there are more ways of making money.
That lack of complexity is another reason so many web companies choose the Apache license when open sourcing their code. "The trend is driven in large part by companies that want to integrate open source development into their fundamental open source project strategy, but don't want that headaches that arise from the GPL," Radcliffe says. In other words, they don't have to worry about giving up the code they don't want to give up.
The overarching theme is that open source software is mixing freely with proprietary code. That's what the Apache license allows. Olson sees it as the future, pointing to giants such as Oracle and IBM that have built successful businesses around open source software projects. "I think that successful open source businesses will look a lot more like IBM or Oracle than Red Hat or MySQL or Sleepycat," he says.
And he believes that even without licenses such as the GPL in place, companies will continue to contribute to open source projects because they now understand how valuable this can be. "As the industry has matured, I think that people have internalized the value of collaboration," he says. "You don't need to browbeat them."
Facebook and Yahoo and Twitter have taught the world. And so has Mike Olson.
