CampSmalltalk: Smalltalk and XP

One of the biggest mistakes I've seen Smalltalk groups make is to take a heavy, large-group, approach. One group I talked to had designers in the US, coders in India, and testers back in the US. There were well over a hundred involved. I heard later that when push came to shove, a small group of the designers did their own coding and testing and got it to run. I wasn't surprised.

In a large project, the big problems are communications between people. The language you use has little effect. Smalltalk makes the biggest difference on small projects, where people focus on communicating with each other and writing good code, and don't have to worry about writing everything down so that people in other time zones can read it. When used correctly, it can make a tremendous difference. I know of two projects that built similar systems in a similar time-frame, but one has 100 people and the other had a dozen. From what I can tell, the large group had at least a dozen people at least as good as anybody on the other team. They were slowed down by communication overhead.

To me, XP seems like the perfect software development method for Smalltalk. It doesn't seem to scale, but everything has its limits, including Smalltalk.

XP prevents many of the errors I have seen Smalltalk developers fall into, including analysis paralysis, being too smart and building something too complicated, and treating Smalltalk like OO Cobol.

I think that Smalltalk would have been even more successful than it was if XP had been written down ten years ago.