As a microbiologist who has published work based on in-house computer code, I empathize with the laments in your News Feature (Nature 467, 775–777; 2010). However, its emphasis on more training for scientists and students fails to consider a downstream issue — retaining the talent afterwards.
From personal experience, there is a common perception that computational analysis is just a tool to enable 'real' discoveries on the bench. Hence, there is little incentive to brush up on computer-related skills. A few graduate students from my laboratory have invested time and effort to become respectable programmers, despite their lack of a background in computer science, but they became disillusioned when they found that their hard work was considered to be of secondary importance to the science.
These talented individuals have since left for careers in finance, management and information technology, where the same programming know-how and problem-solving skills are highly appreciated. They now enjoy shorter hours, comparable pay and greater job security than a tenure-track assistant professor.
The corollary of Nick Barnes's observation in World View that “most professional computer software isn't very good” (Nature 467, 753; 2010) is that good programmers — regardless of their scientific background — are in demand everywhere. Teaching programming skills and best practices to scientists may indeed improve the quality of software, but the software in question might not necessarily be for scientific research.