OSCON 2009: Lying and Geniuses

Alas, it's taking me much longer to write up my OSCON trip report than I had planned. Part of that is that I've been busy with other things, like resurrecting a testbed for the SCM framework, so that I can start on the workaround for the "unwanted mounts" issue. And part of it is that it's just taking me longer to do the writing, despite already having an outline. So since I have finished my comments on a couple talks, I'll go ahead and post those now.

My favorite session at OSCON was "How to Lie Like a Geek" by Michael Schwern, which was a lively talk about things we can do to help or hinder communication. Most of us are familiar with technical lying, particularly bad benchmarking and abuse of statistics. Michael recommended shootout.alioth.debian.org as a place to get good benchmarks.

Michael also suggested several ways to "lie" that are more generic. Many of these had to do with getting so caught up in details that the big picture is lost. For example, there's lying by information overload: providing too much detail. There's also lying by pedantry or by being excessively literal: focusing on the wrong details.

A fascinating way to mislead is to state the obvious. For example, consider this exchange (which I've adapted from the example Michael used):

Alice: We should change this code.
Bob: But that could introduce bugs.
Alice: Of course. Why are you arguing against making the change?
Bob (annoyed): I'm not, I'm just saying we might introduce bugs. Don't put words in my mouth.

I've seen exchanges like this in the past. At best, there's this unnecessary hiccup in the conversation. At worst, the conversation goes off into the weeds, with endless rounds of claims and counter-claims about who said what and what was really meant.

When I mentioned this example to my wife, she pointed out that people often make implicit requests in their statements. So if Bob says something obvious, he's clearly not offering new information, so it's not surprising for Alice to interpret his comment as a request. If Bob had a different request in mind, like "let's defer that until after the code freeze", it'd be more helpful to say that in the first place.

Another talk that I liked a lot was "Programmer Insecurity and the Genius Myth", by Ben Collins-Sussman and Brian Fitzpatrick. I had to chuckle at this quote, which Ben and Brian said came from the 2008 Google I/O developer conference:

Can you guys make it possible to create open source projects that start out hidden to the world, then get "revealed" when they're ready?

The opensolaris.org site supports hidden projects. And for awhile we created new projects as hidden, so that the project team could get the project page set up before making the project visible. But hiding the project doesn't provide that much benefit. After all, the project space is a work area; it's okay for things to be rough. And we've had a few problems with projects lingering in hidden mode for a long time. So we've moved towards making projects visible from the start.

Of course, it's natural for people to want to get things right before sharing them with the outside world. But there are potential advantages to sharing early. One is that if you're going down the wrong path, early sharing improves the odds that you'll discover the problem quickly. And if you have to explore a couple dead-ends before you come up with something that works, the information from those explorations is available for others to learn from.

And then there's the "bus factor", which refers to the number of people who would have to all be run over by a bus to stop the project. Working in the open--with archived discussions, publicly visible code, any plans and documentation that you've written down--makes it easier to recover if a team member is unavailable for whatever reason.

Ben and Brian did point out that it is possible to share too early. The project needs to be far enough along that it won't get stalled when outsiders show up and start asking questions and suggesting changes.

But since working in the open like this can be challenging, what are things we can do to encourage it? Ben and Brian made several suggestions, both social and technical.

On the social side, the first thing to remember is not to let your ego get tied up in your design or code. You've probably heard that before, but I think it bears repeating. When someone points out a problem in my design or code, I try to think of it as an opportunity to improve something that I care about, and as an opportunity to learn. But that's often not my initial, automatic response; it takes some effort.

Even if you don't expect your project to benefit from outsiders' comments, engaging in a conversation can have benefits. As Ben and Brian put it, if you want to influence people, you need to be open to influence.

On the technical side, consider what behavior your tools are encouraging. For example, the current opensolaris.org portal doesn't keep page histories, which discourages its use for collaborative writing and editing. That's one of the things that will be fixed by the move to XWiki in September.

Ben and Brian also recommended responding to questions and arguments on the project web site, rather than by email. With that approach, they said, the discussion is less likely to degenerate into pointless argumentation. I don't recall them saying why they think this works. I suppose one reason is that it helps keep arguments from being repeated.


Fitzpatrick/Collins-Sussman gave the same talk at Google I/O 2009 apparently, which was recorded:


Posted by richlowe on August 12, 2009 at 08:40 PM PDT #

Post a Comment:
Comments are closed for this entry.

Random information that I hope will be interesting to Oracle's technical community. The views expressed on this blog are my own and do not necessarily reflect the views of Oracle.


« June 2016