Why Good-First-Bugs often aren't

by David Humphrey

Let me start by saying that I'm a huge fan of open source projects putting a Good-First-Bug type label on issues. From my own experience over the past decade trying to get students engaged in open source projects, it's often a great way to quickly find entry points for new people. Right now I've got 40 students learning open source with me at Seneca, and because I want to get them all working on bugs ASAP, you better believe that I'm paying close attention to good-first-bug labels!

Maintainers

There are a few ways that we tend to interact with good-first-bugs. First, we have project maintainers who add the label when they think they see something that might be a good fit for someone new. I've been this person, and I know that it takes some discipline to not fix every bug. To be honest, it's faster and easier to just fix small things yourself. It's also going to mean you end up having to fix everything yourself--you won't grow new contributors this way.

What you need to do instead is to write large amounts of prose of the type "Here's what you need to do if you're interested in fixing this". You need sample code, links to relevant files, screenshots, etc. so that someone who lands in this bug can readily assess whether their current skill level (or aspirational skill level), and the bug's requirements, meet.

Sometimes maintainers opt not to do this, and instead say, "I'd be willing to mentor this." The problem with this approach, in my experience, is that it becomes a kind of debt with which you saddle your future self. Are you sure you'll want to mentor this bug in 2 years, when it's no longer on your roadmap? You'd be better to "mentor" the bug upfront, and just spell out what has to happen in great detail: "Do this, this, and this." If you can't do that, the reality is it's not a good-first-bug.

New Contributors

The second way we encounter good-first-bugs is as someone looking for an opportunity to contribute. I make a habit of finding/fixing these in various projects so that I can use real examples to show my students the steps. I also tag along with my students as they attempt them, and I've seen it all. It's interesting what you encounter on this side of things. Sometimes it goes exactly as you'd hope: you make the fix and the patch is accepted. However, more often then not you run into snags.

First, before you even get going on a fix, finding a bug that isn't already being worked on can be hard. A lot of people are looking for opportunities to get started, and when you put up a sign saying "Start Here," people will! Read through the comments on many good-first-bugs and you'll find an unending parade of "I'd like to work on this bug!" and "Can you assign this to me?" followed by "Are you still working on this?" and "I'm new, can you help me get started?". That stream of comments often repeats forever, leaving the project maintainers frustrated, the contributors lost, and the bug untouched.

Expiry Dates

Once people do get started on a bug, another common problem I see is that the scope of the bug has shifted such that the problem/fix described no longer makes sense. You see a lot of responses like this: "Thanks for this fix, but we've totally refactored this code, and it's not necessary any more. Closing!" This doesn't feel great, or make you want to put more effort into finding something else to do.

The problem here wasn't that the bug was wrong...when filed. The bug has become obsolete over time. Good-first-bugs really need an expiry date. If a project isn't triaging its good-first-bugs on a somewhat regular basis, it's basically going to end up in this state eventually, with some or all of them being useless, and therefore bad-first-bugs. You're better off closing bugs like this and having no good-first-bugs listed, than to have 50 ancient bugs that no one on the project cares about, wants to review, or has time to discuss.

Good First Experience

This week I've been thinking a lot about ways to address some of the problems above. In their lab this week, I asked my students to build Firefox, and also to make some changes to the code. I had a few goals with this:

  • Build a large open source project to learn about setting up dev environments, obtaining source code, build systems, etc.
  • Gain some experience making a change and rebuilding Firefox, to prove to themselves that they could do it and to remove some of the mystery around how one does this.
  • Learn how to start navigating around in large code, see how things are built (e.g., Firefox uses JS and CSS for its front-end).
  • Have some fun doing something exciting and a bit scary

I've done this many times in the past, and often I've gone looking for a simple good-first-bug to facilitate these goals. This time I wanted to try something different. Instead of a good-first-bug, I wanted what I'll call a "Good First Experience."

A Good First Experience tries to do the following:

  • It's reproducible by everyone. Where a good-first-bug is destroyed in being fixed, a good-first-experience doesn't lose its potential after someone completes it.
  • It's not tied to the current state of the project, and therefore doesn't become obsolete (as quickly). Where a good-first-bug is always tied to the project's goals, coding practices, and roadmap, a good-first-experience is independent of the current state of the project.
  • It's meant to be fun, exploratory, and safe. Where a good-first-bug is about accomplishing a task, and is therefore defined and limited by that task, a good-first-experience can be the opposite: an unnecessary change producing an outcome whose value is measured by the participate vs the project.

Toward a Good First Experience with Firefox

I reached out to a half-dozen Mozilla colleagues for ideas on what we could try (thanks to all who replied). I ended-up going with some excellent suggestions from Blake Winton (@bwinton). Blake has a history of being whimsical in his approach to his work at Mozilla, and I think he really understood what I was after.

Based on his suggestions, I gave the students some options to try:

  • In browser/base/content/browser.js change the function OpenBrowserWindow to automatically open cat GIFs. You can alter the code like so:
function OpenBrowserWindow(options) {
  return window.open("http://www.chilloutandwatchsomecatgifs.com");
}

  • Look at the CSS in browser/base/content/browser.css and try changing some of the colours.

  • Modify the way tabs appear by playing with CSS in browser/themes/shared/tabs.inc.css, for example: you could alter things like min-height.

  1. You could try adding a background: url(http://i.imgur.com/UkT7jcm.gif); to the #TabsToolbar in browser/themes/windows/browser.css to add something new

  2. Modify the labels for menu items like "New Window" in browser/locales/en-US/chrome/browser/browser.dtd to something else.

None of these changes are necessary, prudent, or solving a problem. Instead, they are fun, exploratory, and simple. Already students are having some success, which is great to see.

Nature via National Park vs. Wilderness

I was reflecting that the real difference between a good-first-experience and a real bug is a lot like experiencing nature by visiting a National Park vs. setting out in the wilderness. There isn't a right or wrong way to do this, and both have obvious advantages and disadvantages. However, what National Parks do well is to make the experience of nature accessible to everyone: manicured paths, maps with established trails to follow, amenities so you can bring your family, information. It's obviously not the same as cutting a trail into a forest, portaging your canoe between lakes, or hiking on the side of a mountain. But it means that more people can try the experience of doing the real thing in relative safety, and without a massive commitment of time or effort. It's also a mostly self-guided experience vs. something you need a guide (maintainer) to accomplish. In the end, this experience might be enough for many people, and will help bring awareness and an enriching experience. For others, it will be the beginning of bolder outings into the unknown.

I don't think my current attempt represents a definitive good-first-experience in Mozilla, but it's got me thinking more about what one might look like, and I wanted to get you thinking about them too. I know I'm not alone in wanting to bring students into projects like Mozilla and Firefox, and needing a repeatable entry point.