There are 139,834 open source projects under way on SourceForge, the popular open source hosting site. Five years from now, only a handful of those projects will be remembered for making lasting contributions--most will remain in niches, unnoticed by the rest of the world. For every Linux, Apache, or MySQL, dozens of other open source efforts fizzle out.
That's a dilemma for the many companies that are expanding their use of open source. Corporate developers and other IT professionals must get better at divining the winners and ignoring the losers. The wrong picks can lead companies down a rat hole of support problems and obsolete software.
Good bets for the next round of open source innovation include the Mule enterprise service bus, Alfresco content management system, and Spring framework for Java applications. But what about the 139,831 other options?
One promising project on the bubble is OpenVista, software for managing health records and health care operations. OpenVista has a few things going for it: a strong code base, pressing demand, and a company bent on commercializing it. Yet OpenVista shows how leadership fissures that a proprietary software company might work through can become deep rifts in open source projects.
OpenVista was posted on SourceForge on June 6. It wasn't a big surprise; the posting had been promised several times by Medsphere, a company founded to commercialize OpenVista. But things unraveled quickly. In three weeks, Medsphere sued co-founder and CTO Steve Shreeve, who was responsible for the posting. In a complaint filed in Superior Court of Orange County, Calif., Medsphere charged that Shreeve and his older brother, Scott, Medsphere's chief medical officer at the time, had breached their fiduciary duty as directors, violated confidentiality agreements, and caused the company to suffer $50 million in damages. CEO Ken Kizer and board members contend that the Shreeves should have held a meeting before posting OpenVista to review what code would be released.
Steve Shreeve responds that parts of the source code had been released two times previously without any meetings with the CEO or board of directors. "I'm the largest shareholder in the company," he fumes. The Shreeve brothers were fired in June after the blowup, saying they were only taking steps that are commonplace in nurturing an open source project, like creating a foundation around it.
Kizer argues the suit "is not about open source code. It's about corporate governance." He maintains the Shreeve brothers, after a falling-out with Kizer, tried to use open source as a cover for launching a separate venture. Steve Shreeve denies such an intent.
Now, without the Shreeves, Medsphere is again planning to launch OpenVista on SourceForge. Kizer's hope: "This will be a big event in health care, and a community will be attracted to OpenVista."
But Steve Shreeve thinks Medsphere has burned a bridge, destroying that intangible element of trust that
leads people to work on open source projects, as paid contributors, volunteer coders, or interested users offering vital suggestions. "How is any open source developer going to have a shred of confidence to work on this?" asks Shreeve. "How can they convincingly say they've released the code when they're suing me over its release?"
If building open source code requires tactful leadership, short lines of communication, and trust among team members, then OpenVista comes to the open source market with one hand tied behind its back. But those aren't the only factors businesses must assess. A review of successful open source projects, and insights from those who've led and implemented them, reveal common themes and metrics that can be used to measure a project's potential for success.
This story was updated Feb. 26.
Tony Wasserman thinks the success factors are so clear that he can assign numbers to them. Wasserman, director of the software management program at Carnegie Mellon West, a Silicon Valley branch of Carnegie Mellon University, is developing a business-readiness rating service for open source code. SpikeSource (an assembler of open source code stacks), publisher O'Reilly Media, and Intel are helping with the project.
The Business Readiness Rating service is collecting public feedback on its proposal for evaluating open source code. Eventually, says Wasserman, it will host automated software tools that harvest statistics from open source project sites that help predict their likelihood of success: the number of developers and core developers, frequency of releases, support queries and unanswered queries, and the number of bugs tracked versus fixed. Those metrics will then be used in a decision-making framework to sort through open source projects. Some automated evaluation tools already exist. They include FLOSSmole (the Free/Libre Open Source Software Mole), which automatically burrows into data on an open source project site such as page views, downloads, bandwidth consumed by downloading, and number of comments posted.
Such tools have their place, but IT pros still will need to make judgment calls as open source proliferates beyond the Linux operating system and Apache Web server.
Amazon.com, Google, The Hartford, and Sabre Holdings are some of the companies that have benefited from betting early on the right open source programs. Sabre, for example, bet on MySQL for quickly serving travel information to consumers. The circle of companies willing to place their own bets continues to grow. Social networking site hi5 Networks, which gets about 18 million unique visitors a month, manages its Linux-based systems using a new piece of open source software called Hyperic HQ. As a 35-employee company, hi5 has a small IT staff and hadn't invested in a commercial system management product such as BMC Patrol or CA Unicenter.
Hi5 had used another piece of open source code, Big Brother, but decided Hyperic HQ offered more, such as alerts when systems dip below performance thresholds and historical performance analysis. As a user of Apache, Linux, and the open source PostgreSQL database, CTO Akash Garg was inclined to "give open source systems management a chance." Why? For the same reason many other companies are drawn to open source. "It's a lot cheaper," he says.
Hi5 tested the quality of the Hyperic community by posting questions to it, and had problems getting adequate answers at first. It got around that by building relationships directly with the support team for Hyperic Inc., the company behind the project, which was eager to work with an early customer of an emerging project. Hyperic has been downloaded more than 26,000 times since it became available as open source code last summer; a commercially supported version is also available from Hyperic Inc. for $500 a year.
Keeping a strong community is going to be the challenge for the Apache Harmony project, which was organized to generate an open source version of Java. IBM, which wants to keep Java as an open system despite Sun Microsystems' ownership, was a major backer. But Sun in November made its Java code open source, sucking the air out of Harmony's sails. Harmony will continue, says its chairman, Geir Magnusson, but what passionate user community will form around Harmony when open Java is available on the Net?
In contrast, JBoss has succeeded so well as an open source Java application server that there's scarcely any oxygen left for competitors. Before buying JBoss for $350 million last April, Red Hat said it would distribute and contribute to the ObjectWeb Jonas project, based in France. Now it's "Jonas who?" even at Red Hat. The user community had voted. Jonas may have its adherents in Europe, but in North America, it was wheezing. Separately, there are questions about whether the innovative Geronimo application server project at Apache will be able to run in JBoss' wake or just limp, even with support from IBM, which bought a company, Gluecode, that included Geronimo in its open source software stack.
Harmony and Jonas have one thing in common: They drew not on grassroots developer or user support for a core idea, but on high-level support from interested companies--IBM and Red Hat-- with their own agendas. When priorities changed for those companies, confidence in the sustainability of the project faded among supporters.
Fedora Legacy Linux is another project that foundered without a passionate user following. Linux-enthusiast developers started the Fedora Legacy project with a goal of providing security and bug fixes for older versions of Red Hat Fedora Linux. Unfortunately, the larger community didn't share a handful of developers' enthusiasm for "legacy" Linux, and the project was closed down at the end of last year because of the lack of support.
An active community is part of what set Apache apart in the mid-1990s from other freeware, of which IT managers were rightfully wary, says Apache founder Brian Behlendorf. Instead of a site packed with free code, at Apache.org potential business users found the code as it was being developed, with comments being exchanged on recent work. "It was easy to ask questions, to sign up for the mailing list, to see the long conversational threads on support questions," he says.
Behlendorf is describing the transparency that still marks any vibrant open source project. A community needs to be measured by its activity and transparency as much as its size. The reasons for decisions must be clear, with threads of discussion in forums leading up to them, and negative and positive comments getting their airing. That's one of open source's most powerful ingredients. "It's hard inside a company to say what you really think. Outside, people can afford to be brutally honest," says James McGovern, chief security architect with The Hartford's property and casualty IT group.
A healthy community also polices itself, jumping on those given to excesses, such as personal attacks. Strong forums and mailing lists focus on the big picture and don't obsess on the small stuff, including the personality differences. That's where leadership enters the calculation.
As a Web site designer for Organic Online, Behlendorf proved he had a knack for putting out the right question to the right crowd--for example, how to improve the kludgy Web server from the National Center for Supercomputing Applications--and for recognizing merit in the responses. Other programmers developed the Apache code, but Behlendorf hosted the project, using wit and firmness in steering what should go in and what should stay out. He promoted Apache's merits at industry events and public forums.
MySQL, Linux, and other successful open source projects all have this in common: a Linus Torvalds sort of figure, a benevolent dictator with the humility to see the value in other people's work. At JBoss, it was Marc Fleury. At MySQL, it's a pair of developers, Monty Widenius and David Axmark, who produced the early versions of MySQL and selected the smooth Marten Mickos as CEO. Ross Mason is the undisputed development leader of Mule, an enterprise service bus gaining traction at financial institutions. Mason's also the founder of MuleSource, the company behind it. At Samba, founded in 1992 to provide file and print capabilities across Windows, Unix, and Linux, it's the diplomatic yet decisive Jeremy Allison.
Successful projects are characterized by long incubation periods with delayed rewards. Apache took four years and Subversion "five years of heads-down work," says Behlendorf. It wasn't clear at the start that either Apache or Subversion would pay off, he says. It's the job of the benevolent dictator to keep a team together through the long march, to impose discipline, assign work, award praise, and heal rifts created by setbacks.
At fractious Medsphere, CEO Kizer is allied with board member and former CEO Larry Augustin, who was angered by the posting of Medsphere's code. Both vie for OpenVista leadership with Steve Shreeve, who left medical school before graduation to become Medsphere's founding CEO. Shreeve later moved into the CTO role, and in his comments, there's a subtext of worrying about how Augustin and Kizer lay the blame for project delays at his feet, despite what he saw as extraordinary effort. The pairing of long gestation with few rewards in the early going breeds tensions that can easily devolve into infighting. Medsphere's OpenVista needs a benevolent dictator to manage the tension.
Rod Johnson, originator of the Spring Framework, is a central figure for that project, one who represents another hallmark of successful open source projects: a public champion who embodies the ideas of the project, even when core developers remain at home writing lines of code.
JBoss built aspect-oriented programming into the application server. The thinking behind the move: Something that's used over and over in an application, such as identity management or security, should be an aspect of the program and not have to be repeated with variations in different parts. So JBoss came up with a tagging method that became so successful that the Java Community Process followed its lead and built it into Java. Now it's all the rage to avoid Java's complexities and substitute simpler technologies.
When it first came out, MySQL was a lightweight database, lacking many of the features of Oracle or IBM's DB2. But it did one thing superbly: function as an ultrafast, read-only database that could serve HTML pages in high-traffic situations. MySQL was adopted by startups for that purpose, then caught on with larger Web businesses such as Travelocity. It has since filled in many of its database technology gaps--but it still serves pages fast.
Linux is a version of Unix, yes, but its innovation was to tailor Unix to commodity hardware without losing the operating system's strengths. None of the big Unix vendors seized on that concept. Sun's initial Solaris for x86 was a crippled system that didn't catch on, leaving an opening for Linus Torvalds. The power of Linux lies in its ability to fully exploit cheap hardware instead of being dumbed down.
Apache could take easy-to-write PHP plug-in extensions and use them to connect to different resources, and it was designed to scale up to meet traffic demands. In short, it was an innovation that was just right for the times. It now runs 60% of the active sites on the Web, according to Netcraft's Internet-monitoring site, compared with 31% for the runner-up, Microsoft's Internet Information Server.
Firefox has gained market share by improving the browsing experience while Internet Explorer stagnated. Compare that with, say, OpenOffice, which has been around for years with no significant traction against Microsoft Office. If a piece of open source software doesn't do something exceptionally well, if it's merely a cheap imitation of what you can buy, don't bet on it as the next big hit.
Apache and Subversion are "products that were well timed," contends Behlendorf. "They occurred at the beginning of intense need for their kind of functionality." In other words, they were leading-edge innovations. Zack Urlocker, executive VP of products at MySQL AG, says open source code must overturn an established way of doing things. This is open source's "sometimes-overlooked role as a disruptive force," Urlocker says.
The key question to be asked by corporate developers: Is this open source project solving a problem my colleagues and I are wrestling with? Even OpenVista could overcome its troubles if the demand is great enough. Google's Dibona says OpenVista has been anticipated so long developers will give the code a look. "If the code is what the user base is demanding in software, developers will go ahead and work with it," he says. "I've seen projects launched in troubled circumstances before."
Johnson, leader of the Spring project, saw that pressing need as a developer for financial institutions in London. He thought Java Enterprise Edition had become too complex for mere mortals and was actually slowing down the development process. "I saw project after project start with Java, and it wouldn't get the job done," he says.
Asked if he thinks, like Behlendorf, he could pull off a second open source success, Johnson acknowledges the accidents of timing and thought leadership in a rapidly changing technology scene. "The best open source code almost has the stuff of magic to it, the way it fits a growing need," he says. "And magic doesn't come around that often."