Reproducible research [From the Editor]

One of the concerns of the broader field of computational research is whether the results that are published are reproducible. When the author had a position as a senior research scientist at the National Center for Supercomputing Applications in the late 1990s, he had a conversation with a computational physicist who said that most simulation codes in his field were not made publicly available, and it was his opinion that at least 95% of those codes contained major errors that result in technically incorrect results regularly being published. Because he was a generally balanced and agreeable fellow and had worked for many years at the U.S. Department of Energy national laboratories responsible for the development of most of the foundational computational software used today the author considered his estimate of 95% reasonably accurate. This computational physicist said that the solution to the high occurrence of errors is that all computational codes should be made publicly available so that errors can be found and fixed, and he commented that computational research would progress much more quickly if this simple practice were employed. He later received a major national award based largely on the computational codes that he had developed over his career, all of which he had made publicly available. So the author asks readers: "Should authors in the control field be expected or compelled to make their software public, as a way to reduce errors and to facilitate progress in the field?"