Kowari as a Data Set

My Ph.D. direction was a lot to chew. Although I have made a career out of ignoring people's advice when something was hard, I had to take it this time. I have had many suggestions that my thesis was going to be problematic because it was so interdisciplinary and therefore hard to find places to publish. An anthropolical journal might say that a complex systems journal should publish a given paper, and a complex systems journal might suggest that it belongs in archaeology, and so on. So, I am changing direction and taking the advice to continue work on the fundamental problem of the origin of agriculture following graduation. The new direction is a complex systems analysis of the Kowari codebase. Simple, short, bounded and relatively easy to publish. Whew!

Kowari is interesting for several reasons, the most important of which are:

• It is a large, enterprise product with actual users.


• We have a substantial cvs history for it.


• We have recorded knowledge of what developers were doing during its development.

The knowledge of what happened when is the kicker. Many Open Source projects could claim the first two, but they don't typically have any way to corrolate what developers were doing during the development of the project. That data should allow some interesting analyses to be done, such as how "busy" objects changed during the development lifecycle, the relationship between a scale-free network representation of the code and major changes to the codebase, and whether I can manage to develop an algorithm to represent the increasing complexity of the project over time. That last one is interesting since it may be possible to develop a better metric for evaluating "build vs. buy" decisions than source lines of code counters.

Dave Carrington at UQ will become my associate supervisor and Simon Kaplan (now Dean of the Faculty of Information Technology at QUT) will remain my primary supervisor. I will stay enrolled at UQ.

Watch this space for the results.