The CHAD meeting was not all that well attended, but I thought it was worthwhile being there. The speaker, David J. Anderson, did not really follow his slides. Instead, he told stories about his own experiences in software engineering, starting as a game author in the eighties and getting around to how FDD evolved.
He also improvised a model, Coad style, on the topic we chose, which turned out to be horse racing. It was interesting because he was able to almost apply a template that made sense and seemed to take into account a lot of things that people often forget. The whole idea that generic models can be applied to most situations is deeply appealing.
I took down just one note. Estimates should follow a power law.
You know how it's usually done. You list all your tasks, trying not to forget anything. You estimate each one. You add it all up. I've even written spreadsheets that try to automate some of this, with mixed results.
I had not thought too much about the relationship between difficulty levels. However, humans tend to think on linear scales. This extends to estimating: if we think something is a bit more difficult, we assign a bit more time to it. Thing is, software is one of those things where a little more difficult
translates to you'll need twice the time
.