Open Source Founders Reflect On Project Milestones

We met with Wireshark founder Gerald Combs at the annual Wireshark conference in Los Altos, California. His story starts while working at an ISP a few years ago and unfolds with his on-the-job need for an inexpensive protocol analyzer. Other than TCPdump, there were only proprietary network analyzers on the market at the time. With his employer’s approval, Combs architected and developed the code for an open source network analyzer. It was soon downloaded hundreds of thousands of times.

Combs’ cites the project’s rapid popularity as the initial milestone. Over the following months, he received emails asking about his plans for the project. He realized it needed his full-time commitment. His employer at the time had little interest in an open source network analyzer but eventually Combs found the right partner. Today CACE Technologies is the project sponsor, a change he cites as the second significant milestone — as well as CACE becoming his employer.

A third milestone was the build-up of the user and developer community. “We attracted people who stuck with Wireshark — people smarter than me”, he claims. “And we needed an infrastructure including CVS and mail lists that could leverage their talents.” Combs explained that most Wireshark contributors work for larger companies than CACE. The parallel architecture of Wireshark makes their participation easier, letting them work simultaneously on their own pieces, such as protocols.

The Asterisk Project

Mark Spencer, founder of Digium and the Asterisk project, was in San Francisco for the Open Source Business Conference in April and gave us his thoughts. As Spencer explains, Asterisk was started from a real business requirement — his own company needed a phone system. “Telephony” he recalls, “was a large market with no Microsoft-like monopoly, and back then phone systems were expensive.”

To keep a complex development project like Asterisk running smoothly, Spencer set up an issue tracker, something he still considers a significant milestone. Asterisk has many contributors to manage, he explains. He initially targeted a technical audience with Asterisk, adding further, that he designed the code base to support “extreme customization”.

Spencer likewise encouraged the developers to earn income from the Asterisk ecosystem. Digium sponsored the first Asterisk conference to provide them an opportunity to do commercial business and discuss how to move the project forward.

The Mondrian Project

The MySQL Conference in Santa Clara offered an opportunity to speak with Julian Hyde, founder of the Mondrian project. Mondrian is an OLAP cube that Hyde wrote as a skunk project with his company’s permission. Mondrian, he explains, is based on an OLAP standard he created — one which is similar to the Microsoft OLAP product.

Mondrian’s popularity took off when an industry analyst highlighted it in an article. “Open source” Hyde tells us, “provided a multiplier effect, allowing people with the same interests to converge on one project.” But the milestone was when people started to use it, he adds. Hyde explains that using SourceForge was not a watershed but it helped get visibility. Awareness attracted Andreas Voss, for example, a developer with whom Hyde agreed to provide cross-support, giving Mondrian a technical boost through JPivot’s navigational capability.

Even with Mondrian’s success, Hyde didn’t want to start a company. Instead he looked for someone to back the project. He met Richard Daley of Pentaho at a conference, and was impressed with their business model and professionalism. Hyde now has two roles there: architect of the roadmap and community advocate. “Engineers at other companies using Mondrian are full-time contributors” he explains, and “these people have a strategic interest in stable Mondrian code.” Having a commercial sponsor helped the project achieve maturity and industry strength, including regular releases, Hyde says.

Hyde is looking ahead to another milestone. He would like the open source community to help adopt OLAP4J, because, he explains, “the commercial world has failed to provide us with a standard.”