Communication is one of the most underestimated of the XP values. The value of communication is not underestimated, so much as how hard it is. There are many different ways to believe we are truly communicating, but they so often fall short. When we are talking about software development, we have even more communications models to consider.
One of the most obvious is face to face, just plain talking to each other. I still feel that this is the best, most effective way to communicate. This is why co/located teams are always the preferred mechanism for agile development. Not only are we able to have a real/time understanding of what someone is saying, we are able to pick up on subtle body language and nuance that may or may not come out in the written form. When communicating face to face, we still need to be aware of what we are saying and how we are saying it. We still need to do our best to remember the context around the conversation, and ensure we are actually hearing what is being said. I have many times had a conversation where I reflected back what I thought I had heard, only to have the speaker say "no, that is not what I said." This is common enough that I encourage everyone to avoid assuming they understood something, but to actually confirm that they did.
I am definitely a fan of the written word. Even so, it has its drawbacks. To a certain extent, when I write something, I am talking *at* you, not *with* you. There is no space for refinement, at least not instantaneously. It's a little better with tools like IM, which is at least qausi/real time, but still falls short on nuance. In the end we discover that for actual communication, written documents are far less effective than any other form.
So what about those forms of communication that are so important to agile software development? How do we communicate with our stake holders, customers, and development teams? So often we choose some sort of document. We write up a very large requirements document, from which we create a specification document, which will then have a design document attached. Finally, we will write some code and create the working software.
Along the way, we will spend a lot of time going over these documents. We will have meetings to discuss them. We will create drafts that we will meet over and refine. Only when we are finally convinced that we have a particular document right will we move to the next step in our documentation. Now, I've been to a lot of these meetings, and it always turns out the same way. If we allocate, say an hour, to the meeting, the first 45 minutes will be discussing whether or not we have used the correct format, filled in the right blanks at the right time, and generally done everything we can to satisfy the documentation standard. The last 15 minutes, if we can, we will talk about the actual content.
At some point in this crazy process, someone finally writes some code. NOW we are getting somewhere. Well, maybe not. Code is also a form of written communication. If we write hard/to/understand or very complex code, we end up once again getting into a place where communication suffers. We then create a lot of comments that will "explain" what we are trying to do. At least at the moment we wrote the code. Code changes though. As we start working with the actual software, we see things we would like to either modify, or perhaps even fix. Most of the time this happens in a hurry, so we change the code, and don't have the time to ensure the comments are updated as well. In essence, we start to lie to future readers. "This code does this, and it does it this way" only lasts until the first revision.
This all brings me to my main point about communication. Once we've written all of our documents, argued over the formats, written the code, reviewed the code (now *there* is an area where you can really see arguments about format) and delivered the code, what are we really going to be discussing? The *working software*. We look at what it is doing. We play with it. We talk about how to make it better, faster, stronger. We don't go back to discuss the document ever again.
So let's spend less time worrying about the document. In Extreme Programming, and of course in many of the other agile methodologies now, we talk about fitting a user story on an index card. Any further elaboration comes best from writing software and talking about what we have. Another great way of communicating our needs is to write automated acceptance tests. We are identifying how the software is expected to work and we are unambiguously showing what we want.
So to circle back to where we began, communication is about shared understanding. Whatever mechanism we find most useful, we should embrace. We have found that the most succesful mechanism, without exception, is the spoken word. If for whatever reason we find that to not be possible, maybe due to the distributed nature of some agile teams, we need to look for the best way to communicate and overcome the inherent difficulties in not being able to talk to each other directly. The best way to communicate what the software should do is to provide automated acceptance tests. And of course the best way to communicate what the software does is to create it and use it.