[not Rp] Reproducibility of 'Poincare dodecahedral space parameter estimates'
This highlights reproducibility challenges in scientific computing, even with open data and software, which is an incremental but important issue for researchers in computational sciences.
The paper attempted to reproduce a 2008 cosmic topology study that used public data and free software, but found that updating the Fortran 77 code with modern compilers was too difficult, making the results less reproducible than expected.
Is a scientific research paper based on (i) public, online observational data files and (ii) providing free-licensed software for reproducing its results easy to reproduce by the same author a decade later? This paper attempts to reproduce a cosmic topology observational paper published in 2008 and satisfying both criteria (i) and (ii). The reproduction steps are defined formally in a free-licensed git repository package "0807.4260" and qualitatively in the current paper. It was found that the effort in upgrading the Fortran 77 code at the heart of the software, interfaced with a C front end, and originally compiled with g77, in the content of the contemporary gfortran compiler, risked being too great to be justified on any short time scale. In this sense, the results of RBG08 are not as reproducible as they appeared to be, despite both (i) data availability and (ii) free-licensing and public availability of the software. The software and a script to reproduce the steps of this incomplete reproduction are combined in a new git repository named 0807.4260, following the ArXiv identity code (arXiv:0807.4260) of RBG08.