In my culture: the responsibilities of open source maintainers

post by philh · 2020-04-13T13:40:01.174Z · LW · GW · 3 comments

Contents

  When is responsibility taken on?
  If you have responsibility, what does that entail?
  What if someone fails to act responsibly?
None
3 comments

If you maintain an open source project, what responsibilities do you have towards your users? Some recent drama (that I won't link to) reminded me that there are large differences in how people answer that question.

(In the drama in question, this wasn't the only thing at issue. But it was a relevant question.)

I thought I'd take a stab at describing my best guess as to how we answer it in my culture: in the culture that exists only in my mind, but that I think (and hope) many copies of me would implement, if we had that opportunity. (That culture doesn't attempt to exclude people who aren't copies of me, so it does need to be robust to attack. In my culture, we do not just assume in defiance of all evidence that everyone is friendly and trustworthy.)

Some of this will probably seem obvious to many readers, like "in my culture, murder is considered bad". Probably different bits to different readers. I'm interested in discovering which bits seem obvious to almost everyone, and which bits are controversial.

A lot of it follows from how I think about responsibility in general. But if you start to think "extending this response to this other situation, you'd get this, and that's a terrible idea"… in my culture, we don't immediately assume from this that I'm endorsing a terrible idea. Instead we check. Maybe I disagree [LW · GW] that that's how it extends. Maybe I hadn't thought about this, and you can change my mind about the initial response. Maybe I just straightforwardly endorse a terrible idea: in that case, it'll be much easier to push back once you've gotten me to admit to it.

I do not intend, in this essay, to discuss whether any particular person or group is living up to the standards I outline here. I may do that in future [LW · GW]. But how likely that is, and what that follow-up looks like, depends on whether the responses to this essay suggest a lot of people agree with my culture.

I think there are at least three important limitations to this essay. One is that I've never been a maintainer of an open source project that had users other than myself (that I knew of), though I've made small contributions to a few. As such, I don't really know what the experience is like or how my culture deals with its challenges, much like I don't really know how my culture deals with the challenges of living in Antarctica. I can make educated guesses, but that's all they are. I'm not going to explicitly flag them as such in the main body. (I'm already using the words "in my culture, we…" far too much. I'm not changing it to "in my culture, my educated guess is that we…") Ideally, I wouldn't write this essay because there was already a conversation taking place between people who knew what they were talking about. Unfortunately, as far as I've seen there mostly isn't, so I offer what I can.

Another is that I don't talk at all about the responsibilities of users, which is also an important part of the question. I'd like to, but… this essay has been knocking around in my head for at least a year and a half, I've made at least one previous attempt to write it that I gave up on, and I'm worried that if I don't publish it quickly I never will. I hope and believe that even with this omission, it is better for this essay to be published than to not be published.

(I also omit the question "what is this responsibility thing anyway?", but that seems less important to me right now. I've probably also overlooked important parts of the things I do talk about, but that too may not be a big deal.)

And a third is that without specific examples, what I've written is less constrained than what's in my head. It may well be possible to read this and agree with what I've written, and then to discover that your culture has a much stricter or much looser conception of responsibility than mine does.

With all of that out of the way: there are three questions I'm going to be focusing on here. When is responsibility taken on; what does it entail; and how do we react if people fail at it?

When is responsibility taken on?

In my culture, taking on responsibility requires active involvement, but not explicit consent. It's possible to take on responsibility through negligence - to draw responsibility towards yourself without realizing that's what you're doing. In my culture, we're sympathetic towards people who do that, but we don't consider that this absolves their responsibility.

In my culture, merely making something available does not make you responsible for it. If you want to put something online and ignore anyone who tries to use it, you can do that. You are allowed to shithub.

In my culture, you take on more responsibility for a project if…

A common thread here is that responsibility is related to justified expectations. "Expectations" in both the moral sense and the probabilistic sense. If someone can make a compelling argument "this person morally ought to accept responsibility here", or a compelling argument "I predict based on past behaviour that this person will act as though they've accepted responsibility here", then in my culture, that person has some degree of responsibility whether they like it or not.

Accordingly, in my culture you can disclaim responsibility in advance, simply by saying that you're doing this. Note that a pro forma warranty disclaimer in your LICENSE file isn't sufficient here. Instead you should say it at the entry points to your project - probably the README and the website (if there is one). Something like…

Note that while I think this project will be useful to many people, I am not devoting much attention to it, and I'm prioritizing my own use cases. Bugs may be fixed slowly if at all. Features may be removed without warning. If this isn't acceptable to you, you should probably not rely on the project; or at least you should be willing to fork it, if necessary.

Or even:

Note that while I think this project will be useful to many people, I have no interest in accepting responsibility for other people's use of it. Feel free to submit issues and pull requests; but I may ignore them at whim, and my whim may be arbitrary and capricious.

This won't usually be necessary. But if you feel in danger of taking on more responsibility than you'd like, you can do this.

If you have responsibility, what does that entail?

In my culture, if we've taken on responsibility and now we want to discharge it, we do so with care. We give advance notice, and we try to find replacement maintainers. If we can't find a replacement maintainer, we still get to quit. But we try.

In my culture, we acknowledge that different people have different needs and goals, and not every project will be suitable for all users. We try to help users figure out whether our projects will be suitable for them before they get invested. We're open about the limitations of our projects, and about the design goals that we explicitly reject.

In my culture, we don't need to reply to every critic. But we do try to notice common threads in criticism, and address those threads. ("Address" doesn't mean that we necessarily try to solve them in a way that will satisfy the critics. It simply means we give our thoughts about them. We try to do that in such a way that even if the critics aren't happy with what we're doing, they at least feel like we've heard them and understood them.)

In my culture, we accept that some critics will be unkind, unfair and unproductive. This sucks, but it's a reality of life right now. We distinguish these critics from others. We are free to ignore them, and to ban them and delete their posts from the spaces we control. We do not use their actions to justify shirking our responsibilities to other critics.

In my culture, we also take care to check whether we're unfairly rounding people off as being unkind, unfair and unproductive. We don't require perfection from our critics. We try to hold ourselves and our supporters to the standards we expect of our critics. We moderate our spaces, but we try not to silence all criticism from them.

In my culture, we still introduce bugs, because we're still human. We try not to, of course. But we accept fallibility. When we fail, we do our best to fix it. We put more effort into avoiding more catastrophic bugs.

In my culture, we do not accept incompetence or indifference. These are signs that a person should not have taken on responsibility in the first place. We expect people to know the limits of their ability and of how much they care.

In my culture, we can set boundaries around how our future actions are constrained. We distinguish public APIs (where we mostly try to maintain backwards compatibility) from private APIs (where we mostly don't, and if people expect us to we point to the word "private"). We may reject bugfixes that we expect to cause too much ongoing future work, in hope of finding a better fix in future.

In my culture, sometimes we hurt people deliberately because the alternatives are worse. When we do, we own it. For example, sometimes we decide to remove a feature that people were relying on. We don't make that decision lightly. When we do it, we explain why we decided to do it, and we apologize to the people who were relying on it. We don't try to minimize their pain. We try to suggest alternatives, if we can; but we try not to pretend that those alternatives are any more suitable than they actually are. We give advance notice, if we can. We can be convinced to change our minds, if new information comes to light.

Of course, the degree to which we feel bound by all of this depends on the degree to which we've actually taken on responsibility. And note that in all of this, there's very little requirement for positive action. If we don't want to include a feature, we don't need to write it, or even merge the PR if someone else writes it. If we don't want to go in the direction our users want us to, we don't have to go there. We just have to make it clear that that's what we're doing.

What if someone fails to act responsibly?

That is, if someone has taken on responsibility, whether intentionally or not, and then has failed to act like someone who's taken on responsibility… how do we respond?

In my culture, we mostly respond by acting as though this will happen in future.

We don't treat this as an indelible black mark against all aspects of the person. We'll still accept their contributions to other open source projects, for example. (Though perhaps not technically complex contributions, that will likely need a lot of maintenance from the same person going forward.) We'll accept their conference talks. We'll stay friends with them.

But we won't rely on their projects, because we don't expect their projects to be reliable. We warn others about this, but if others decide to use their projects anyway, we consider that none of our business.

We accept that people have a right to run projects as they see fit. We acknowledge that our conception of "responsibility" is a mere social convention. But social conventions are important. We provide an opt-out for this particular social convention, but if you don't opt out explicitly, we'll make it explicit for you.

We give second chances. If they want to take on responsibility in future, and if they seem to understand how they failed in the past, and they say they'll try to do better, we give them a shot, because that's reason to think it won't happen in future. If they still fail, and want a third chance, we take their second failure into account too.

We may fork a project, even if the maintainer doesn't approve, but we'll rarely attempt a hostile takeover from the outside. We'll make sure our fork is clearly distinguished from the original project.

  1. To the extent that this isn't true, I think most of the responsibility comes from something other than "being an open source maintainer". For example, if you're making a web scraper, I would suggest you have a responsibility to make it well-behaved by default - sleep between requests, stop after a certain number of failures, and so on. But that responsibility comes from the product, not the process. You can't avoid it with a disclaimer. And so it's not the focus of this essay. 

3 comments

Comments sorted by top scores.

comment by Dagon · 2020-04-13T20:30:45.714Z · LW(p) · GW(p)

I like this framework of "in my culture", and I mostly subscribe to the specifics you lay out. I might quibble with some of the terminology - "responsibility" can mean a few different things, and you haven't laid out the rewards that are contingent on discharging one's responsibility well.

You mostly describe responsibility as something that happens based on your actions (and on your users' reactions). I think that's fine, but I'm not sure how responsibility differs from consistency. If you've never maintained backward-compatibility in an API, people won't expect it and it won't be seen as irresponsible.

Replies from: philh
comment by philh · 2020-04-15T17:17:13.535Z · LW(p) · GW(p)

Yeah, I sort of hinted at what I mean by responsibility but it would probably be good to make explicit. I'm less confident about what responsibility-in-my-culture is than what it involves, but to take a stab at it anyway...

I mostly think of it as a social attempt to solve a coordination game. Broadly speaking, it's good for people to be able to do the things they want; and in some situations, it's good for people to be able to predict what other people will do. One way to solve this is to say "people can just do what they want as long as it doesn't infringe on anyone's rights" but I think that doesn't give very good results. ("Rights" is another rabbit hole here. Not going there.) I think better is to say "people can just do what they want as long as it doesn't infringe on anyone's rights, but also people are able to make commitments about future behaviour", and if they violate those commitments they get punished in some way. My conception of responsibility goes slightly further and says... "here is a set of default expectations we have about people's behaviour. We think they're not too constraining. In some cases, if you don't like them, you can opt out and do what you want. But if you're going to do that, you need to opt out explicitly, and people are at liberty to take that into account in their interactions with you". And then if someone violates one of those expectations, they get punished in some way.

you haven’t laid out the rewards that are contingent on discharging one’s responsibility well.

Yeah, this didn't occur to me at all. Thinking about why... I think to me, the rewards beyond "not getting punished" are mostly much more individual than the responsibilities are. If you maintain an open source project, you presumably get some form of reward for it - maybe prestige, maybe the warm glow of helping people, maybe you're getting paid. It'll be different for different people. But it doesn't really matter what they are. If you can get the rewards without taking on responsibility, good for you. If you can't, then you either take the rewards with the responsibility, or neither. If you try to take the rewards but avoid the responsibility, that's a problem for society-at-large: you're giving people incorrect assumptions about your future actions, in a way that transfers value from them to yourself, and in a way that (mildly) undermines the whole social edifice we've got built up. And so we discourage that, and try to make you choose, and try to give people more correct expectations.

Replies from: Dagon
comment by Dagon · 2020-04-15T20:13:50.976Z · LW(p) · GW(p)
the rewards beyond "not getting punished" are mostly much more individual than the responsibilities are

I think this is key to any theory of responsibility. Especially when the "punishment" section is also pretty anemic.

In my culture, we mostly respond by acting as though this will happen in future.

If you act irresponsibly, we will expect you to ... act irresponsibly.

I suspect there _IS_ a huge value in trust of one's peers and users, and receiving respect on mailing lists and conferences. There's no formal tracking or clear "you get it if you take responsibilty, you lose it if you fail in a responsibility", but it's still a primary reward/punishment channel about responsibility.