Avoiding Common Pitfalls

Don't Post Without a Purpose

A common pitfall in online project participation is to think that you have to respond to everything. You don't. First of all, there will usually be more threads going on than you can keep track of, at least after the project really gets going. Second, even in the threads that you have decided to engage in, much of what people say will not require a response. Development forums in particular tend to be dominated by three kinds of messages:

  1. Messages proposing something non-trivial

  2. Messages expressing support or opposition to something someone else has said

  3. Summing-up messages

None of these inherently requires a response, particularly if you can be fairly sure, based on watching the thread so far, that someone else is likely to say what you would have said anyway. (If you're worried that you'll be caught in a wait-wait loop because all the others are using this tactic too, don't be; there's almost always someone out there who'll feel like jumping into the fray.) A response should be motivated by a definite purpose. Ask yourself first: do you know what you want to accomplish? And second: will it not get accomplished unless you say something?

Two good reasons to add your voice to a thread are a) when you see a flaw in a proposal and suspect that you're the only one who sees it, and b) when you see that miscommunication is happening between others, and know that you can fix it with a clarifying post. It's also generally fine to post just to thank someone for doing something, or to say "Me too!", because a reader can tell right away that such posts do not require any response or further action, and therefore the mental effort demanded by the post ends cleanly when the reader reaches the last line of the mail. But even then, think twice before saying something; it's always better to leave people wishing you'd post more than wishing you'd post less. (The second half of Poul-Henning Kamp's "bikeshed" post, referenced from the section called “The Softer the Topic, the Longer the Debate”, offers some further thoughts about how to behave on a busy mailing list.)

Productive vs Unproductive Threads

On a busy mailing list, you have two imperatives. One, obviously, is to figure out what you need to pay attention to and what you can ignore. The other is to behave in a way that avoids causing noise: not only do you want your own posts to have a high signal/noise ratio, you also want them to be the sorts of messages that stimulate other people to either post with a similarly high signal/noise ratio, or not post at all.

To see how to do that, let's consider the context in which it is done. What are some of the hallmarks of an unproductive thread?

  • Arguments that have been made already start to be repeated in the same thread, as though the poster thinks no one heard them the first time.

  • Increasing levels of hyperbole and involvement as the stakes get smaller and smaller.

  • A majority of comments coming from people who do little or nothing, while the people who tend to get things done are silent.

  • Many ideas discussed without clear proposals ever being made. (Of course, any interesting idea starts out as an imprecise vision; the important question is what direction it goes from there. Does the thread seem to be turning the vision into something more concrete, or is it spinning off into sub-visions, side-visions, and ontological disputes?)

Just because a thread is not productive at first doesn't mean it's a waste of time. It might be about an important topic, in which case the fact that it's not making any headway is all the more troublesome.

Guiding a thread toward usefulness without being pushy is an art. It won't work to simply admonish people to stop wasting their time, or to ask them not to post unless they have something constructive to say. You may, of course, think these things privately, but if you say them out loud then you will be offensive—and ineffective. Instead, you have to suggest conditions for further progress: give people a route, a path to follow that leads to the results you want, yet without sounding like you're dictating conduct. The distinction is largely one of tone. For example, this is bad:

This discussion is going nowhere. Can we please drop this topic until someone has a patch to implement one of these proposals? There's no reason to keep going around and around saying the same things. Code speaks louder than words, folks.

Whereas this is good:

Several proposals have been floated in this thread, but none have had all the details fleshed out, at least not enough for an up-or-down vote. Yet we're also not saying anything new now; we're just reiterating what has been said before. So the best thing at this point would probably be for further posts to contain either a complete specification for the proposed behavior, or a patch. Then at least we'd have a definite action to take (i.e., get consensus on the specification, or apply and test the patch).

Contrast the second approach with the first. The second way does not draw a line between you and the others, or accuse them of taking the discussion into a spiral. It talks about "we", which is important whether or not you actually participated in the thread before now, because it reminds everyone that even those who have been silent thus far still have a stake in the thread's outcome. It describes why the thread is going nowhere, but does so without pejoratives or judgements—it just dispassionately states some facts. Most importantly, it offers a positive course of action, so that instead of people feeling like discussion is being closed off (a restriction against which they can only be tempted to rebel), they will feel as if they're being offered a way to take the conversation to a more constructive level, if they're willing to make the effort. This is a standard the most productive people will naturally want to meet.

You won't always want a thread to make it to the next level of constructiveness—sometimes you'll want it to just go away. The purpose of your post, then, is to make it do one or the other. If you can tell from the way the thread has gone so far that no one is actually going to take the steps you suggested, then your post effectively shuts down the thread without seeming to do so. Of course, there isn't any foolproof way to shut down a thread, and even if there were, you wouldn't want to use it. But asking participants to either make visible progress or stop posting is perfectly defensible, if done diplomatically. Be wary of quashing threads prematurely, however. Some amount of speculative chatter can be productive, depending on the topic, and asking for it to be resolved too quickly will stifle the creative process, as well as make you look impatient.

Don't expect any thread to stop on a dime. There will probably still be a few posts after yours, either because mails got crossed in the pipe, or because people want to have the last word. This is nothing to worry about, and you don't need to post again. Just let the thread peter out, or not peter out, as the case may be. You can't have complete control; on the other hand, you can expect to have a statistically significant effect across many threads.

The Softer the Topic, the Longer the Debate

Although discussion can meander in any topic, the probability of meandering goes up as the technical difficulty of the topic goes down. After all, the greater the technical complexity, the fewer participants can really follow what's going on. Those who can are likely to be the most experienced developers, who have already taken part in such discussions many times before, and know what sort of behavior is likely to lead to a consensus everyone can live with.

Thus, consensus is hardest to achieve in technical questions that are simple to understand and easy to have an opinion about, and in "soft" topics such as organization, publicity, funding, etc. People can participate in those arguments forever, because there are no qualifications necessary for doing so, no clear ways to decide (even afterward) if a decision was right or wrong, and because simply outwaiting other discussants is sometimes a successful tactic.

The principle that the amount of discussion is inversely proportional to the complexity of the topic has been around for a long time, and is known informally as the Bikeshed Effect. Here is Poul-Henning Kamp's explanation of it, from a now-famous post made to BSD developers:

It's a long story, or rather it's an old story, but it is quite short actually. C. Northcote Parkinson wrote a book in the early 1960'ies, called "Parkinson's Law", which contains a lot of insight into the dynamics of management.

[...]

In the specific example involving the bike shed, the other vital component is an atomic power-plant, I guess that illustrates the age of the book.

Parkinson shows how you can go in to the board of directors and get approval for building a multi-million or even billion dollar atomic power plant, but if you want to build a bike shed you will be tangled up in endless discussions.

Parkinson explains that this is because an atomic plant is so vast, so expensive and so complicated that people cannot grasp it, and rather than try, they fall back on the assumption that somebody else checked all the details before it got this far. Richard P. Feynmann gives a couple of interesting, and very much to the point, examples relating to Los Alamos in his books.

A bike shed on the other hand. Anyone can build one of those over a weekend, and still have time to watch the game on TV. So no matter how well prepared, no matter how reasonable you are with your proposal, somebody will seize the chance to show that he is doing his job, that he is paying attention, that he is here.

In Denmark we call it "setting your fingerprint". It is about personal pride and prestige, it is about being able to point somewhere and say "There! I did that." It is a strong trait in politicians, but present in most people given the chance. Just think about footsteps in wet cement.

(Kamp's complete post is very much worth reading, too; see bikeshed.com.)

Anyone who's ever taken regular part in group decision-making will recognize what Kamp is talking about. However, it is usually impossible to persuade everyone to avoid painting bikesheds. The best you can do is point out that the phenomenon exists, when you see it happening, and persuade the senior developers—the people whose posts carry the most weight—to drop their paintbrushes early, so at least they're not contributing to the noise. Bikeshed painting parties will never go away entirely, but you can make them shorter and less frequent by spreading an awareness of the phenomenon in the project's culture.

Avoid Holy Wars

A holy war is a dispute, often but not always over a relatively minor issue, which is not resolvable on the merits of the arguments, but where people feel passionate enough to continue arguing anyway in the hope that their side will prevail. Holy wars are not quite the same as bikeshed painting. People painting bikesheds may be quick to jump in with an opinion, but they won't necessarily feel strongly about it, and indeed will sometimes express other, incompatible opinions, to show that they understand all sides of the issue. In a holy war, on the other hand, understanding the other sides is a sign of weakness. In a holy war, everyone knows there is One Right Answer; they just don't agree on what it is.

Once a holy war has started, it generally cannot be resolved to everyone's satisfaction. It does no good to point out, in the midst of a holy war, that a holy war is going on. Everyone knows that already. Unfortunately, a common feature of holy wars is disagreement on the very question of whether the dispute is resolvable by continued discussion. Viewed from outside, it is clear that neither side is changing the other's mind. Viewed from inside, the other side is being obtuse and not thinking clearly, but they might come around if browbeaten enough. Now, I am not saying there's never a right side in a holy war. Sometimes there is—in the holy wars I've participated in, it's always been my side, of course. But it doesn't matter, because there's no algorithm for convincingly demonstrating that one side or the other is right.

A common, but unsatisfactory, way people try to resolve holy wars is to say "We've already spent far more time and energy discussing this than it's worth! Can we please just drop it?" There are two problems with this. First, that time and energy has already been spent and can never be recovered—the only question now is, how much more effort remains? If some people feel that just a little more discussion will resolve the issue in their favor, then it still makes sense (from their point of view) to continue.

The other problem with asking for the matter to be dropped is that this is often equivalent to allowing one side, the status quo, to declare victory by inaction. And in some cases, the status quo is known to be unacceptable anyway: everyone agrees that some decision must be made, some action taken. Dropping the subject would be worse for everyone than simply giving up the argument would be for anyone. But since that dilemma applies to all equally, it's still possible to end up arguing forever about what to do.

So how should you handle holy wars?

The first answer is, try to set things up so they don't happen. This is not as hopeless as it sounds:

You can anticipate certain standard holy wars: they tend to come up over programming languages, licenses (see the section called “The GPL and License Compatibility” in Chapter 10, Legal Matters: Licenses, Copyrights, Trademarks and Patents), reply-to munging (see the section called “The Great Reply-to Debate” in Chapter 3, Technical Infrastructure), and a few other topics. Each project usually has a holy war or two all its own, as well, which longtime developers will quickly become familiar with. The techniques for stopping holy wars, or at least limiting their damage, are pretty much the same everywhere. Even if you are positive your side is right, try to find some way to express sympathy and understanding for the points the other side is making. Often the problem in a holy war is that because each side has built its walls as high as possible and made it clear that any other opinion is sheer foolishness, the act of surrendering or changing one's mind becomes psychologically unbearable: it would be an admission not just of being wrong, but of having been certain and still being wrong. The way you can make this admission palatable for the other side is to express some uncertainty yourself—precisely by showing that you understand the arguments they are making and find them at least sensible, if not finally persuasive. Make a gesture that provides space for a reciprocal gesture, and usually the situation will improve. You are no more or less likely to get the technical result you wanted, but at least you can avoid unnecessary collateral damage to the project's morale.

When a holy war can't be avoided, decide early how much you care, and then be willing to publicly give up. When you do so, you can say that you're backing out because the holy war isn't worth it, but don't express any bitterness and don't take the opportunity for a last parting shot at the opposing side's arguments. Giving up is effective only when done gracefully.

Programming language holy wars are a bit of a special case, because they are often highly technical, yet many people feel qualified to take part in them, and the stakes are very high, since the result may determine what language a good portion of the project's code is written in. The best solution is to choose the language early, with buy-in from influential initial developers, and then defend it on the grounds that it's what you are all comfortable writing in, not on the grounds that it's better than some other language that could have been used instead. Never let the conversation degenerate into an academic comparison of programming languages (this seems to happen especially often when someone brings up Perl, for some reason); that's a death topic that you must simply refuse to be drawn into.

For more historical background on holy wars, see catb.org/~esr/jargon/html/H/holy-wars.html, and the paper by Danny Cohen that popularized the term, ietf.org/rfc/ien/ien137.txt.

The "Noisy Minority" Effect

In any mailing list discussion, it's easy for a small minority to give the impression that there is a great deal of dissent, by flooding the list with numerous lengthy emails. It's a bit like a filibuster, except that the illusion of widespread dissent is even more powerful, because it's divided across an arbitrary number of discrete posts and most people won't bother to keep track of who said what, when. They'll just have an instinctive impression that the topic is very controversial, and wait for the fuss to die down.

The best way to counteract this effect is to point it out very clearly and provide supporting evidence showing how small the actual number of dissenters is, compared to those in agreement. In order to increase the disparity, you may want to privately poll people who have been mostly silent, but who you suspect would agree with the majority. Don't say anything that suggests the dissenters were deliberately trying to inflate the impression they were making. Chances are they weren't, and even if they were, there would be no strategic advantage to pointing it out. All you need do is show the actual numbers in a side-by-side comparison, and people will realize that their intuition of the situation does not match reality.

This advice doesn't just apply to issues with clear for-and-against positions. It applies to any discussion where a fuss is being made but it's not clear that most people consider the issue under discussion to be a real problem. After a while, if you agree that the issue is not worthy of action, and can see that it has failed to get much traction (even if it has generated a lot of mails), you can just observe publicly that it's not getting traction. If the "Noisy Minority" effect has been at work, your post will seem like a breath of fresh air. Most people's impression of the discussion up to that point will have been somewhat murky: "Huh, it sure feels like there's some big deal here, because there sure are a lot of posts, but I can't see any clear progress happening." By explaining how the form of the discussion made it appear more turbulent than it really was, you retrospectively give it a new shape, through which people can recast their understanding of what transpired.