The Tool Built Out of Spite and Deadline
For three years, the Linux kernel relied on a proprietary tool that many in the open source community were uncomfortable with.
BitKeeper was proprietary software, and Linus Torvalds knew it. But it was fast, handled thousands of patches, and allowed developers to work in a distributed way that the free alternatives of the early 2000s struggled to match. So the kernel community made a pragmatic deal: BitMover, the company behind BitKeeper, would let them use it for free, under certain conditions.
The arrangement worked well enough, until 2005.
That year, Andrew Tridgell, the Australian programmer best known for creating Samba, started developing a free tool that could communicate with BitKeeper repositories. BitMover's CEO, Larry McVoy, considered this a violation of the software's licensing terms and withdrew free access for Linux kernel developers.
Tridgell saw things differently. During a conference demonstration, he showed how BitKeeper's server exposed protocol information through a simple Telnet connection and a single command: "help." He argued that this was enough information to build a compatible client without reverse-engineering BitKeeper's internal implementation.
The two sides never agreed on what had happened. What followed, however, was clear: the Linux kernel community had lost access to the tool it had relied on to manage its development history.
Linus looked at several existing alternatives. By his standards, none offered the performance or workflow he needed. They were too slow for a project the size of the kernel, and none provided the combination of distributed development, efficient merging, and data integrity he was looking for.
His requirements were straightforward: speed, strong safeguards against data corruption, and the ability to support thousands of developers working in parallel without getting in each other's way.
So, on April 3, 2005, he started writing his own.
He announced the project on April 6. By the following day, Git was already capable of managing its own source code. On April 16, it was used to import the Linux kernel's 2.6.12-rc2 source tree, covering more than 17,000 files and 6.7 million lines of code. Just two days later, the first merge of multiple branches took place.
The speed of development was remarkable. In less than two weeks, Linus had built something that was already handling the kind of work it was designed for.
He named it "git," a British slang word for an unpleasant or foolish person. The name was a joke at his own expense. In the project's original README, he explained his naming philosophy:
"I'm an egotistical bastard, and I name all my projects after myself. First Linux, now git."
Even the manual page described it, with typical humor, as the "stupid content tracker."
Linus wasn't the only developer responding to the BitKeeper crisis. Around the same time, Matt Mackall began developing Mercurial, another distributed version control system. Both projects emerged from the same moment of uncertainty, although Git would eventually become the more widely adopted of the two.
But perhaps the most interesting part of the story came next.
Linus never intended to spend years maintaining a version control system. He had built Git because he needed a better way to manage the Linux kernel, not because he wanted another long-term project to look after.
So, on July 26, 2005, he handed maintenance of Git to Junio Hamano, one of its major early contributors. Hamano took the project forward, released version 1.0 that December, and has remained its principal maintainer ever since.
That handoff is easy to overlook, but it may have been just as important as the initial development. A tool born out of frustration still needed someone willing to maintain it, refine it, and build a community around it. Linus supplied the initial direction, and Hamano helped turn it into a lasting project.
Today, Git has become the dominant version control system across the software industry. Every time you commit, branch, or merge, you're using a tool whose origins can be traced back to a withdrawn license, a disputed protocol, and a developer who refused to wait for someone else to solve his problem.
The lesson isn't that you should build everything yourself. It's that constraints, even frustrating ones, can force you to rethink existing solutions and create something better than what comfortable conditions might have produced.
Enjoy reading on PBLISHD?
Bring your reading experience closer. Install PBLISHD on your device.