You may have heard of the Joel Test to evaluate startups:
- Do you use source control?
- Can you make a build in one step?
- Do you make daily builds?
- Do you have a bug database?
- Do you fix bugs before writing new code?
- Do you have an up-to-date schedule?
- Do you have a spec?
- Do programmers have quiet working conditions?
- Do you use the best tools money can buy?
- Do you have testers?
- Do new candidates write code during their interview?
- Do you do hallway usability testing?
At Taodyne, we pass all these, yet our project took a major blow because there is at least one other key rule that goes above all these. I will call it rule zero, and it impacts how you write code.
- Do you go in a single direction?
The code-oriented formulation of rule zero is the following:
- Do you avoid branches like the plague?
Do not ever fork your codebase
With tools like Git, Mercurial or Bazaar, branches are much easier to manage than they were in the past. They actually became part of the process. With these distributed source control systems, you keep creating and merging branches for practically everything you do.
So why the title, “Branches are evil” (a favorite of über-geek Dennis Handly)? Here, by “branch” I don’t mean the kind of minor branching where two variants of the codebase differ by a few lines of code. This has to happen as you develop code. The problem is when the fork affects the design itself, or the direction the project is following.
My biggest mistake at Taodyne was ignoring this key rule. It all started with a simply thought experiment about an alternative design, quickly followed by a little bit of experimental code. And I forked our code. But then, others started building useful stuff on top of that branch.
When we decided to drop that alternate design, we had a big problem on our hands. We needed to salvage all the good stuff on that branch, that had been written against a completely different design. We have spent several months trying to reconcile these two conflicting designs that had conflicting objectives. As a result, I spent countless hours trying to fix bugs that were a result of merging code with different assumptions.
When a single mistake delays the release of your product by a few months, you know you did something solidly wrong. I thought I’d share the experience so you won’t do the same mistake.