As the scandal over the climate model at the University of East Anglia reminds us, much cutting edge science relies on computer modeling. Professors working in such fields as econometrics, population genetics, and climate science routinely write their own programs or extensively modify existing programs to do the work they want. This raises serious problems about the results they publish, because, as we all know by this point in the information age, all programs have bugs. Some sample studies of custom academic software suggest that much of it is riddled with major errors. Because of this, many journals now require that copies of any custom software used by researchers be submitted along with articles for publication. But not all do, and even when the programs are submitted, who is going to check them line by line? Some climate modeling programs are a hundred thousand lines long.
The President of the National Academy of Sciences, Ralph Cicerone, recently wrote an editorial in Science calling for greater scientific openness, in particular the sharing of all data. But if the software used for modeling is not also shared, the data in some cases is not of much use.
Watch for this to be a major issue for science funding bodies over the next decade.