Mediaboss Marketing

Monty Meets Monty

Monty Taylor says he wound up on the MySQL project because he shared a name with Monty Widenius. Widenius is the Finn who founded MySQL the worlds most popular open source database and Taylor says that when he applied for a job at MySQL AB, the company that oversaw the project, it was his name that elevated his resume above the slush pile. How could they not interview someone else named Monty? he says, with the sort of playfulness that so often laces his storytelling.

He came to the project in a roundabout way. As a college undergraduate, he dabbled in computer science, but it rubbed him wrong. My first professor looked like Ed Grimley, he says. All my worst fears of what computer science would do to me were realized. In the end, he studied theater at a tiny Texas school called Abilene Christian, and after graduating, he continued to work as a director and a lighting designer and all-around backstage technician. He drifted back into software only because he needed another way to pay the bills and a local Fujitsu office needed a system administrator. But at MySQL, even as he continued his work in the theater, he cemented his place in a world that would remake the software business.

MySQL was a stepping stone for the software game but also for Monty Taylor. Following in the footsteps of Linux, the project built up a vibrant community of developers, but this community splintered after tech giant Sun Microsystems acquired MySQL AB in 2008. Though Sun was ostensibly a friend of open source software, its commercial aims clashed with the project. MySQL never reached the point where it transcended commercial contraints where it was free to evolve as quickly as possible.

At Sun, in an effort to solve this problem, a developer named Brian Aker forked the MySQL project, creating a new version called Drizzle, and Taylor soon joined the effort. The Drizzle team eventually moved Rackspace, and though the project never revived the success of MySQL, it brought Taylor to the doorstep of OpenStack.

With Rackspace working hand-in-hand with NASA on the project, the idea from the beginning was that OpenStack would work like Linux. No one company would have control. The rub was that software development moves so much quicker on the web than it does with operating systems. When he left the Drizzle project for OpenStack, Taylors task was to build a service that could not only meld contributions from any number of developers spanning any number of companies, but meld them with unusual speed and accuracy.

The result was the OpenStack Continuous Integration service though Taylor doesnt like calling it that. For one thing, its a dreadful name. But it also fails to show that the OpenStack system a very different from the norm.

Built around tools like Jenkins and Buildbot, CI systems are designed to rapidly merge new code into a large piece of software. But this doesnt always happen as quickly as it could. Typically, new code is tested only after it has been merged into the trunk, the core software the community is working on. This means that, at any given time, the trunk may be broken, and for a sweeping project like OpenStack, a broken trunk puts a drag on the process.

What happens is that people trying to get work done start pulling broken code from the trunk, Taylor says, and thats not scalable.

OpenStacks CI service is different in that Taylor and others have developed a means of automatically testing all code before its submitted to the trunk. There is no red light, green light, he says. The trunk is always green. It just always works. We dont let change land unless it works. This protects developers from each other, but it also makes it easier for anyone to use the code, at any time. They always know the trunk is in order.

Link:
Why Open Source Software Is Like Burning Man (Only Better)