Beyond training in programming, scientists creating production code need training in software engineering. The folks over at Software Carpentry (http://software

Beyond training in programming, scientists creating production code need training in software engineering. The folks over at Software Carpentry (http://software-carpentry.org/ http://software-carpentry.org/) seem to have the right idea. Just as important, like other forms of scientific experimentation, work based on computer analysis needs to be independently reproducable. A big part of that is to make the source code and data available to other researchers (ideally, to the public at large) and well-documented. It seems like the CRU people failed in both areas.