The trouble with designing these big, complex solutions is that we often then get stuck when trying to actually implement them. They are just too hard and have too many points of failure. These solutions end up becoming the bane of our existence. We have to maintain them and fix bugs in them. Sure, we can be good citizens and use design patterns and well-documented architectures. Ultimately, however, we end up, time and time again, juggling big issues while trying to get a new piece of functionality into the system.

Among the XP practices are some techniques that will help us reduce the complexity of the problem.

Ideally, we should find some way that we can all be geniuses. If we are all geniuses, we can always create simple solutions to all of our problems. Something we all know is that it is easy to create simple solutions to solve simple problems. So to become a genius, all we need to do is break down all of our problems into lots of smaller problems that are easy to solve.

Any problem that I estimate to take longer than four hours to solve is too big for me to feel comfortable with. I want to accomplish at least one thing in a day; if I accomplish more than that, I feel even better. This is a personal decision, but most people I have worked with aim for four hours or less. I have never worked with anyone^[1] who was comfortable with tasks that lasted more than six hours. If I get a task that I estimate will take longer than my maximum four-hour period, I break it down into several smaller tasks. If any of those tasks will take longer than four hours, I break those down again. Something that I have found interesting is that the most effective developers I know break down the majority of their tasks so that they require less than an hour to complete.

Let’s see how this works, beginning with a trivial example before I lead you through a more complex problem breakdown.

Suppose the customer has asked us to develop a piece of software that emulates a calculator on the desktop. So this is our problem: develop a .NET Framework calculator. Easy. So how long will the development take? Do you know immediately? I don’t, but I bet it will take longer than four hours, so I need to break down the problem. Here’s a list of smaller problems:

Can I give time estimates for each of these functions? I feel confident that I can, and I estimate between one and two hours for each function. Some people might ask, “Why so much?” After all, they know that I am an experienced developer and might think that I should be able to put a simple calculator together pretty quickly. I am estimating that this will take me up to a day to complete based on my experience. I am breaking the problem down into even smaller steps, and I will go through that process with you now for the add function.

Notice that I emphasize writing tests. In the next chapter, you will learn why and how to write tests before the code. For now, it is important to understand that it gives me more focus on developing only what is required and nothing extra. These tests define the behavior expected of the object that will be developed.

Also notice that I am including the time it takes to write these tests, to develop the user interface, and integrate the code with a source control system. All these extra things take time that is often not accounted for by less-experienced developers and project managers. These extra things are often the reason so many software projects run late; developers too often fail to account for the “other stuff” that we need to do to write high-quality code.

With less-trivial problems, you will often find yourself asking some tricky questions, and this is where having contact with the customer becomes important. It is hard to break tasks down without being able to get answers to questions you have. Consider the preceding example. I have made an assumption that the calculator will deal with negative numbers. What if negative numbers are not permitted or required by the system? Having the customer present while doing this task breakdown will make this very apparent. The conversation in our eXtreme .NET team might go something like this:

Chris: Why are you testing for adding negative numbers? We don’t want to add negative numbers!

Pete: Panic! What should happen if we get a negative number?

Chris: You should inform the user that negative numbers are not allowed.

Eddie: Okay, change that test to “Test for handling negative inputs, 10 minutes.”

Remember one of the practices discussed in Chapter 1 was the whole team. Here is an example where having the customer as part of the team can make a difference. If you cannot get a customer to be present during a task breakdown session, try to get him or her on the phone or instant messenger. Failing that, try your manger or other developers. At least they may spot some mistakes or incorrect assumptions you have made before you actually commit them to code. You should e-mail your customer the results from each task breakdown session so that they can review them and change their minds or reset the direction.

We return to this scenario later on in this book to write code for some of these tasks.

The first thing to happen is that the customer will provide the development team with a story. A story is a simple description of a piece of functionality that the system needs to perform. In a full XP environment, the team will play the planning game. This involves the customer writing a number of stories on cards (usually index cards). Then the development team will place rough time estimates on each story. The customer then decides which stories to do first. The planning game provides a long-term view of how the project could pan out eventually, but also includes short-term goals for the next few weeks. By repeating the planning game every few weeks, the accuracy of the estimates should get better and you enable the customer to choose the direction of the project on a smaller scale. Figure 3-1 shows a sample story card.

Figure 3-1. An XP story card.

A story can be considered similar (and are often compared) to a UML Use Case. The story should not explain all the details of the functionality. A story provides a rough idea of the scope. Stories are promises to hold conversations between the developers and the customer. This conversation occurs when the time comes to begin implementing the story. The conversation enables the developers to explore exactly what tasks will have to be carried out to satisfy the customer that the functionality described in the story is finished.

The best way to start this is to begin a conversation with your customer. Let’s see how our team does it:

Eddie: What would you like this software to do?

Chris: I need to be able to display the time in different places in the world on my desktop.

Pete: You mean you want clocks, one for each place in the world?

Chris: Yes, but I want to be able to choose which places to display the time for, not just have a clock for every time zone that exists.

Eddie: Okay, how about we start with simply doing one clock for one time zone and then build the system from there.

Chris: Um... well, I want to choose the time zone and have multiple clocks.

Eddie: Yes, I understand that, but we want to get some core functionality for you to see as soon as possible, and then we can add to the software from there.

Chris: Okay, as long as you are aware that we will need more than one clock.

Eddie: Sure, so what shall we put down as the first story? How about “Select time zone to display?”

Chris: Sounds good. I also want to be able to put my own label next to the clock, so if I select GMT, I want to be able to label it as London.

Pete: Fine, so the second story could be “Add custom label to selected time zone?”

Chris: Yes that’s right; so what next? I can select a time zone and associate a label with it, but I need to be able to see the clock. Do you need a story for displaying the clock?

Eddie: You’re right; how do you want it displayed? I assume that just a digital text output is okay? Something like “Time in London is: 14:23.” How does that sound?

Chris: No, no, no! You’ve misunderstood me! I need to see a clock. You know, with hands and a round face! Like this. (The customer grabs a scrap of paper and scribbles the picture shown in Figure 3-2.)

Figure 3-2. Customer drawing.

Pete: Panic! That wasn’t what I was thinking.

Eddie:Sure, we can do that, but it will take a bit more work.

Chris:How much work? Doesn’t sound that hard to me.

Eddie: Well let’s break it down and I’ll tell you how long it will take.

Chris: I don’t have time for all this; I just want a clock on my screen for different places in the world!

Eddie: Okay, we’ll get back to you soon with some estimates for timeframes.

Chris: Great, when?

Eddie: In the next hour.

Chris: Oh okay, that’s not so bad.

Some of this conversation probably sounds familiar. It is not uncommon for customers (managers) to be short of time and want answers on the spot. As developers, we need to respect that they have other pressures and do what we can to help them. It is important not to let their stress get to us and force us to make on-the-spot decisions or provide time estimates based on nothing more than numbers plucked out of the air. Notice how Eddie, who has more experience, steers the conversation. Eddie is guiding both Pete and Chris through the process in a way that gives them the best chance of succeeding.

From the preceding conversation, Eddie and Pete can draw the following story cards:

From these stories, could you give reasonable time estimates? For the first two, you may be able to get reasonably close (apart from one issue). The third one is a little trickier because the clock has to show the correct time and be updated as the time changes. Let’s see what Eddie and Pete come up with when they break these stories down.

Eddie has put a question mark next to the first task because neither he nor Pete are sure what is involved in getting all the time zones from the operating system. They know the time zones must be there because when they set the Windows clock they can select from a list of available time zones. To find out where this list is stored, they need to carry out some investigation and experimentation.

In XP terminology, this is called doing a spike or spiking. This spike will be another task. Because they don’t know how long the spike will take, they set an upper limit to the amount of time spent spiking before providing feedback to the customer. Eddie and Pete now have a list of tasks that looks like this:

They can now break down the next story.

Finally, they can tackle the third story: Draw a clock for the time zone and a label for it. Eddie thinks this story is too big, so he proposes they break it down into smaller stories:

From these five stories, they can create a set of tasks for which it should be reasonably easy to provide time estimates. The tasks for each story are estimated as follows.

Based on this task breakdown, Eddie and Pete can get back to Chris, their customer, with some confidence in the estimated timeframes. The timeframes are still estimates, but the developers are happier that they have thought about the job and the tasks to complete. They can also explain that the first thing they need to do is spend half a day understanding how time zone information is stored in the operating system. When they know how the time zones are stored, they can use them in the program.

As you can see, the estimated times for each task are small; all of them except the spike are less than 30 minutes. This is a good indicator that they have got their tasks to the correct level of granularity. If you have several tasks that are four or five hours long, you should think harder about how to break them down. For me, the spike task is still cause for concern. After the customer has had time to review the task breakdowns and give the go ahead on the project, I would break down the spike task further. Let’s see how Eddie and Pete go about doing this.

To break down the spike task, Eddie and Pete need to think about the approach to take to find the information they are after. There are some likely candidates to find out how they can access the time zone information:

Based on the fact they have given themselves 4 hours to come up with some answers, Eddie suggests spending no longer than 45 minutes investigating each area. Because both he and Pete are working on the project, they could each take a different area to investigate. This gives Eddie and Pete a task list for the spiking tasks.

If they come to a dead end before 45 minutes, they can stop and move on to the next area of investigation. Likewise, if they find a mechanism that provides access to the time zone data, they don’t need to continue with any of the other spiking tasks.

Eddie and Pete are now ready to start this project with some very tightly defined tasks and outcomes at the end of each task. This tight definition of tasks enables them to move at a fast and furious pace during the working day.