Do large projects silence osmotic communication?
Or can you still be agile?
Comment Put a bunch of people together in a room and they'll be able to communicate more easily than if they were in different rooms. Walls place barriers between people, after all.
This announcement may have come from the Department of the Bleeding Obvious, but the principle of grouping people together and letting them get on with it is central to the idea of agile software development.
The principle is summed up by OO/use case guru Alistair Cockburn in Agile Software Development :
"It is difficult to build extensive tacit knowledge without good osmotic communication, and that is hard to do with more people than can conveniently fit in a room."
"Osmotic communication" refers to the general drift of information that takes place in a software development room - a couple of programmers discussing a design issue may be overheard by several other programmers, who thus latch onto the latest burning issues in the ongoing project, or a business expert is overheard discussing the problem domain with some developers.
In agile projects it's common to optimise the office layout around open communication. Similarly, prominently displayed diagrams, wall-charts, etc, all help to increase the flow of communication between developers by acting as "information radiators" (a communication-rich office setup isn't without its problems, though. I'll cover these in a future article).
Not that any of this precludes the need for requirements analysis and effective up-front design, of course, but such a setup can certainly help a team work more effectively together. But what happens when the number of people on the project goes beyond a certain size? As soon as you have more developers than can fit in one room, you instantly lose many of the benefits of osmotic communication.
Project size has always been a contentious issue with Extreme Programming . At its outset, XP's creator Kent Beck stated that XP shouldn't even be attempted on large projects, and that the process is best geared towards small teams (up to about 12 people). Since those heady early days, though, many teams have had a go at applying XP on large projects, with varying degrees of success.
The most common trait among these projects seems to be that each team has tailored the process to make it work (not a bad thing, as XP - at least as described in the Second Edition of Beck's book  - is meant to be tailored).
For example, there's a case study of the high and low points experienced by a 50-person XP team  at ThoughtWorks, Inc. (Side note: despite observations such as "code is definitely worse than we started" [sic], their rather surreal conclusion is that "XP, or our evolved version of it, has done wonders for us as a team".)
Osmotic communication isn't the only quality to suffer when a project scales up in team size, but it's certainly the most palpable effect. So it's understandable that the first modifications to the process that XP teams typically make centre on fixing this decay in communication. And the "fixes" that they apply generally involve improving the level of documentation - replacing verbal communication with written. Documentation may be in the form of a project Wiki or emails at one end of the scale, and functional specifications, UML models, use cases etc at the other.
Put another way, if you have a small team of just a few developers and a business expert all in a single room, you may be able to get away with a lack of documentation. But as the project scales up, or as the team becomes more distributed, written documentation quickly becomes essential, even for XP teams.
I would be interested to hear about the success stories and problem points that you've experienced in adapting agile projects (not just XP!) to work with more than a roomful of people. ®