It’s pie in the sky to expect today’s scientists to construct portable scripts to reproduce not only their results but their publication quality graphics.
Desirable, sure, but completely unrealistic.
There is a world of difference between getting a nice script going to make your work more productive, and being able to reproduce your own work cleanly, to the hard work of polishing that script so that it’s ready for public distribution.
If a compentent coder can reproduce the analysis provided in the paper by following its prescriptions, then surely that is a sufficient standard of reproducibility.
Last time I looked at the software carpentry course, I though it was a nice intro to software eng, but I struggled to see how a working atmospheric scientist would integrate the material into their workflow. Well intentioned, but too scattered to make an impression on most of the scientists I work with.
I’m a computer science guy who spends much time trying to get the atmospheric scientists I work with to use SVN, to help maintain shared libraries, and to standardise on coding styles. And I think the standard of programming in science leaves a lot to be desired. But the arrogrance of many CS types about this issue is equally annoying.