Stephen has done some work in investigating tools for the improvement of code quality. Thanks to these efforts, we are setting up virtual environments for the simulation; this makes it easy to replicate development on any machine, and reduces dependency complexity and management. We use the virtualenv tool for this.

Also, we are setting up testing infrastructure. In fact, we aim at a high coverage of tests for the functions the code is supposed to do. Of course, for anyone in software development, that means improvement of code stability and quality while facilitating integration, especially when multiple programmers work together (and, like here, not co-located). We think we’ll use nose as a python tool for continuous automated testing.

As a next step, we’ll implement some first core objects. As we favor agile development, we strive to include the user in shaping those core objects. For now, members of the non-technical stream of the project will take the role of the user.

We are quite excited as things now move on.

Granted, simulations are a tricky thing. They can never cover the whole complexity of a complex system – like the economy. For us though, the simulation is not a means for academic exercise or to model as close as possible reality. It would serve the purpose of getting a feeling of the dynamics of such a system and to get a communication vehicle.

Of course, we could easily set up a 3 years monster project to apply the best methodology, consult relevant information and implement a thoroughly elucidated theoretical foundation into simulation software. However, we don’t think we have this time.

We focus on getting a compromise between speed of implementation and sufficiently significant modeling, allowing us to spot trends and tendencies in the system.

Ideally this would provide us with information on how to parameterise the system, to eliminate any obvious fallacies and to feedback into the theoretical framework.

We are currently validating assumptions for the simulation, which we want to setup as an agent based system. On the feasta website, there is a thread discussing the assumptions for the simulation.