Why Open Source Works (Alex Gaynor)

Transcript

1. Why Open Source Works Alex Gaynor PyCon Canada Sunday, August 11, 13 Hi everyone. Thanks for coming out. I’m super thrilled to be here, this now makes 4 PyCons, in 4 countries, on 3 continents in the last year for me. So that’s really cool, thanks for having me. A huge round of applause for the organizers, this has been a phenomenal conference thus far.
2. Why this talk? Sunday, August 11, 13 So first things first, why give this talk? A lot of people would say “isn’t it enough that it works?” Why do we care why it works?
3. 3 years ago... Sunday, August 11, 13 3 years ago, almost to the day, August 15th, 2010, I gave a talk at an education conference titled “Open Source and Education”. I talked to a group of educators and researchers about the practices we were using in open source. Specifically about the overlaps between techniques from pedagogy, from teaching, and practices we’ve organically developed in the open source community.
4. The cobblers children have no shoes Sunday, August 11, 13 And a few months ago I get approached about speaking here, and I start thinking “what are the things I know about that people might want to hear”. And I realized through conversations I’ve had, a lot of people actively working in open source don’t really know these things. So I figured it was something worth sharing.
5. What do I know about Open Source? Sunday, August 11, 13 Before I really dive in, I figure you might want to know how I know stuff about open source. . I write a lot of it; I’m a core dev for Django, CPython, PyPy; my day job is now working on open source, from libcloud to openstack, and many other projects at Rackspace. I help shepherd it: I serve as a director of the Python Software Foundation, and before that of the Django Software Foundation. And of course, I use it all the time, so I’m approaching this from a lot of angles. Anything I say about psychology, pedagogy, or sociology is something I read in a paper somewhere. I’m not an expert in those fields, please call me out if I get something wrong. The conference organizers seem to be messing with my head a little, they put me between two individuals who are actually qualified to speak on this stuff.
6. What practices are we concerned with? Sunday, August 11, 13 Before we dive into the “why”, we should spend a few minutes discussing the “what” of open source, because open source describes a pretty wide range of behaviors
7. • Code dumps • Public development •Open development Sunday, August 11, 13 So the way I see it there are about three types of software that match the definition of open source. The first is code dumps, here’s some software, look it’s open! This is basically the same as closed source development, except there’s some code you and I can look at at the end. The next is public development, this is something like Android, theres a public VCS repo, you can see people working on it, but the progress isn’t driven by a process open to the public. And finally there’s open development, this is what we’re going to talk about today, and this is probably what most of the OSS projects you know are: this is Django, this is Python, this is OpenStack, this is Linux, etc.
8. • Pull requests / contributing patches • Code review • Bug reports • IRC / mailing lists Open development practices Sunday, August 11, 13 These are some of the practices exhibited by open development projects, certainly not an exhaustive list.
9. • Writing code • Reviewing Code • Providing support Sunday, August 11, 13 Many many many developer hours. Often from either our free time, or our employers time. Why do we volunteer our time like this, why do our employers let us spend *their* time on this stuff? Those are the questions I want to try to answer for us. Because intuitively it’s bonkers, who lets their employees give away their code, what person goes home and does the same type of stuff as they do at the job.
10. What do we believe? Sunday, August 11, 13 So to begin unpacking this, first I wanted to take a look at what people believe themselves. So I put out a survey, 3 questions, “Do you contribute to Open Source?”, “Why do you contribute to Open Source?”, “Why don't you contribute to open source any more (or as much as you'd like to)?”. 150 responses. ~50% contribute regularly, 28% have contributed in the past, 15% have not contributed before, 5% other. Decent sample size, population is biased, it’s all people within one or two degrees of me socially, and I have weird social circles. So those are the opinions I collected.
11. Why do we contribute to open source? Sunday, August 11, 13 There were some really varied answers to “Why do you contribute to open source?”, I’ve tried to coalesce ones that were the same, and I’m only going to talk about some of the more common answers.
12. Because I gained / To give back Sunday, August 11, 13 This was, by a decent margin, the most common answer. A lot of people feel their professional lives have benefited hugely from their usage of open source, so they want to pay it forward, so to speak.
13. Improve a thing I use Sunday, August 11, 13 “Scratch your own itch” is another strong motivator, people who use software hit bugs or need more features and then they want to fix them and share that with the world. If you hang around the development lists for Django or Python you’ve probably heart someone say something like, “Well, I don’t have this need, but maybe if you do and you write the code, it’s worth having”, so people’s intuition has some merit, we write the software we need.
14. For fun Sunday, August 11, 13 A lot of us apparently like writing software enough that we’ll do it in our spare time, as well as for a job.
15. Develop skills / Career advancement Sunday, August 11, 13 I’ve bunched these together, but they’re actually pretty different. People who wrote about developing skills tended to talk about working with talented programmers, getting really good code reviews, or learning new code bases. In contract, career advancement was about the professional visibility that comes from contributing to open source. That’s not to say it’s a bad thing, just a different thing.
16. Be a part of a community Sunday, August 11, 13 People want to be a part of something. In particular people wrote about being a part of something that they didn’t have access to where they lived.
17. Keep the world running Sunday, August 11, 13 Open source powers nearly everything, to one degree or another, a lot of people want to keep that up. This is in some senses a variation on “To give back”, it also encompass things like “I love knowing that tons of companies use my work”.
18. It’s my job Sunday, August 11, 13 Some people get paid to write open source all day. I’m really lucky to be one of them. There are companies literally paying people to give away software. That’s awesome.
19. But doesn’t this make no sense? Sunday, August 11, 13 This is a question I get a lot from friends and family members. Why are you writing all this software and giving it away, couldn’t you sell it and make some money? I’d characterize this as the naive economists point of view. Yeah, I could sell some of the stuff I write (some of it really has no commercial value of course), but I value a lot of other things as well. The master economist knows that there are things besides money which incentivize.
20. The motivation of individuals Sunday, August 11, 13 Thus far we’ve talked a lot about what people believe about what they do. Now I want to dive into some of what the academic literature says. This isn’t specifically research on open source contribution, but more general research on what motivates us to work and to learn.
21. Mastery Sunday, August 11, 13 The first thing is we want to obtain a skill. We all want to be able to say “I’m good at this”, we want to understand in greater depth our craft. Open source rewards and encourages this. We get to work with experts who’ll review our code, we get to take time with no deadlines to perfect our patch, and to try to learn something new. I want to learn more about unit testing, I can find a project that needs some tests and contribute. Why do the “experts” want to take their time to code review? Because they care about the health of their projects, and new contributors are the lifeblood of that, by helping someone level up their skills you’re grooming a future maintainer, and you never know who’s going to outpace you and make the project better than you imagined.
22. Autonomy and creativity Sunday, August 11, 13 People want to be in control of what they do. And no one tells you what to do in open source. Some days I work on a compiler, some days I do cryptography (PS: this is scary), some days I work on an ORM or a web framework. I direct myself to projects I’m interested in, to things I care about. I also get to choose how I work on these things, in terms of the aesthetics of code, architectural decisions. I work with others because I value their knowledge and opinions, but I ultimately have direction over how I work on them.
23. Community Sunday, August 11, 13 We want to be a part of something. To work with and learn from others. To help bring new people into the fold. To see a little piece of software grow up and change the world.
24. Things we don’t have Sunday, August 11, 13 There are also some features we don’t have, that a lot of people think help motivate us, but in fact don’t.
25. Rewards Sunday, August 11, 13 There’s almost no extrinsic rewards for working on Django. The biggest reward I can give another developer, is to make them also a core developer, to enable them to contribute more. There’s a ton of research on this subject, simply put, for intellectual, creative tasks, rewards don’t motivate you. In fact they de-motivate you, they replace your intrinsic motivation with extrinsic. Instead of contributing because you want to help the community, you contribute for the reward, and when the reward goes away, so does your motivation. The funny bit, and there’s research that shows this: even when you give people these facts, they still want to believe that rewards work.
26. Coercion Sunday, August 11, 13 I can’t force you to work on something. We sat down to write documentation about security policy for Django, and this is something important right, particularly we wanted to give people a hard timeline, if you report a security issue, we WILL respond within 48 hours. How do you make that guarantee when it’s everyone’s free time? What can you say, oh do this or we’ll fire you from this thing you donate in your free time? Ethical issues aside, when you’re coerced you perform worse, and you care about the quality of your work less. We can make statements like “response in 48 hours” precisely because everyone participates voluntarily, we understand this is important so we make it a priority for ourselves, and that gives way better results.
27. Evaluation Sunday, August 11, 13 We don’t try to rank contributors, we don’t try to give you a score, no “You get a B+ on Django this month”. It turns out there are 3 primary effects of evaluating in this way, a) You become motivated by the evaluation, the grade becomes a reward, b) you care less about the underlying work, and c) you seek the easiest task that gets you good evaluations, you don’t try to challenge yourself, because that might get you a worse grade at contributing to Django. In contrast, open source gives feedback. Here’s how you can improve this patch, here’s how you’re doing a great job helping out on the mailing lists, if you focus on this aspect of your code reviews, you’ll be more effective.
28. The motivations of companies Sunday, August 11, 13 That’s individuals, but what about companies? Surely they could make more money, and have fewer competitors, why give away the secret sauce? But instead companies are some of the biggest drivers of open source, the majority of the work on the Linux kernel is by people paid to work on it, for example.
29. Giving back Sunday, August 11, 13 No seriously. Companies want to give back to. They recognize that when they choose an open source technology, it’s not a static thing, that they need to help promote and grow the community of developers, and that they can help by improving it.
30. Improvements from the community and accelerating growth Sunday, August 11, 13 An open community can improve software beyond what an individual company can do. OpenStack is a really fantastic example of this, the project moves at a lightning pace, they do hundreds of CI runs a day (often hundreds an hour!), with many companies and individuals contributing.
31. Grow a community Sunday, August 11, 13 Companies want to be a part of a community too. Specifically it means things like, you can hire people with existing skill sets, you don’t have to train someone on your custom framework, it’s the hot new thing that everyone knows already.
32. Competition + Cooperation Sunday, August 11, 13 I think the key to recognizing why corporate involvement in open source is that competition and cooperation are not mutually exclusive, and neither is a static mode of behavior. There are things that it’s important we cooperate on, and there are things where competition helps produce better results.
33. Challenges Sunday, August 11, 13 I’ve been super positive so far, but I want to share some of the challenges as well, like Jacob said yesterday, challenge is just another word for opportunity, it’s places we can get better. These are challenges that individuals face in contributing to open source, and challenges that communities face.
34. Time and money Sunday, August 11, 13 This is a big one for people who want to contribute, or want to contribute more than they do. When their work comes out of their free time, it’s pretty strictly bounded by how much real life interferes. Money clashes a bit with what I said about rewards right, does paying people to work on open source change it’s nature? This is a longer conversation, and I don’t have the research to fully back this up, but my experience says there’s a fundamental difference between paying someone for their freetime, and allowing them to make it their full time.
35. Building welcoming communities Sunday, August 11, 13 This is a huge challenge for open source communities. I think in the Python community in particular has done great work here over the past few years. But there’s always room for improvement. We can do an even better job, being welcoming to people of diverse backgrounds, different levels of ability, and different perspectives. Unfortunately one of the most common reasons people identified for not participating in open source was because of jerky maintainers. We need to strive to be a community where that isn’t ok.
36. Feeling of inability Sunday, August 11, 13 Another common reason not to contribute is feeling like you don’t have the skills or the knowledge or the background needed. I’m not totally sure how we solve this, but we need to, because everyone can help. I have over 120 repositories on github. It’s no exaggeration to say half of those are forks I created to fix a single typo. And wearing my maintainer hat I have to say, I love when someone sends me a typo fix patch, every little bit helps.
37. Cliff Gardner Sunday, August 11, 13 Anyone here heard of Philo Farnsworth? He invented television, in Provo, Utah, and then did the first TV broadcast from San Francisco. He was a visionary, and he died broke. But I really want to tell you about his brother in law, Cliff Gardner. Cliff saw Philo’s drawings, and recognized that this was something that was going to be important, so he said, “I don’t really understand the science behind all of this, but it looks like you’re going to need glass tubes, how would it be if I learned to blow glass and made them for you?”, Philo was inventing the cathode receptor, and glass tubes weren’t really a thing you could buy, “This looks important, and I want to help” is what Cliff said. That’s open source.
38. With apologies to Aaron Sorkin Sunday, August 11, 13 Before I continue the metaphor, I want to point out that the majority of that is taken from a TV show written by Aaron Sorkin called Sports Night, and I highly recommend you all watch. Aaron, if you read or hear this, please don’t sue me.
39. Open source = glass tubes Sunday, August 11, 13 Open source is glass tubes. Almost none of us will start the next linux kernel, but we can all learn to contribute. The majority of open source is not creating the next great new project, it’s contributing a missing paragraph to the docs, it’s writing a great bug report with a test case, it’s adding the small missing feature. We contribute because we’ve built our hobbies or our careers on this stuff, and we know it’s important, because it’s fun, because it helps us learn, because it lets us be a part of something. And we can all learn to blow glass tubes.
40. Sprints Sunday, August 11, 13 If I’ve convinced you there’s any merit at all to this whacky open source thing, I’d like to encourage you to join the sprints. There’ll be people from a lot of projects, who want nothing more than to help you get started contributing.
41. Thank you. alex.gaynor@gmail.com / @alex_gaynor https://speakerdeck.com/alex Sunday, August 11, 13 Before we finish, I want to apologize quickly, original versions of this talk came with citations and more statistics to back up some of the studies I mentioned. Unfortunately those are all in a pad of paper back at my apartment in San Francisco, whoops. If there’s any interest at all, reach out to me, on twitter or email, and I’ll get those online. Thank you all for coming and listening, thanks you all for your past contributions to open source, thank you for your future contributions. We’ve got some time for questions now.