The case for open computer programs

Offener Zugang zu wissenschaftlichen Texten, offener Zugang zu wissenschaftlichen Daten – diese Diskussion hat breite Akzeptanz gefunden.
Der Umgang mit wissenschaftlicher Software – letztlich eine Sonderform von Daten – dagegen wird bisher oft nicht näher betrachtet.
Ince et al. haben sich des Themas in Nature kompakt und übersichtlich angenommen und schlagen Lösungswege für die aufgezeigten Probleme vor:

“There are a number of barriers to the release of code. These include a shortage of tools that package up code and data in research articles; a shortage of central scientific repositories or indexes for program code; an understandable lack of perception of the computational problems with scientific code leading to the faulty assumption that program descriptions are adequate (…); and finally that the development of program code is a subsidiary activity in the scientific effort.

A modest proposal

An effective step forward would be for journals to adopt a standard for declaring the degree of source code accessibility associated with a scientific paper. A number of simple categories illustrate the idea:

• Full source code: full release of all source code used to produce the published results along with self-tests to build confidence in the quality of the delivered code, as is the case with Perl modules in the CPAN archive, for example (http://cpan.org).

• Partial source code: full release of source code written by the researcher accompanied by associated documentation of ancillary packages used, for example commercial scientific subroutine libraries.

• Marginal source code: release of executable code and an application programming interface to allow other researchers to write test cases.

• No source code: no code at all provided.

This hierarchy of disclosure would alert both the readers and authors of a journal article to the fact that the issue is important and would highlight the degree to which results might be reproduced independently. There remain, however, some potential stumbling blocks, a number of which can easily be resolved using existing facilities.”

Darrel C. Ince, Leslie Hatton & John Graham-Cumming: The case for open computer programs. Nature 482, 485–488 (23 February 2012)
doi:10.1038/nature10836

This entry was posted in Forschungsdaten, Open Access, Publizieren, Zeitschriften. Bookmark the permalink.

Leave a Reply