@davidwilby@nshephard Very interesting read, thank you! A definitely agree that sometimes a pragmatic "second best" approach is all that is available and researchers should strive for that! What are your thoughts regarding Matlab toolboxes? Is it possible to also download old versions of them and re-execute them on modern hardware?
@davidwilby a) i think you could've summarised this article; b) i think "reproducibility within a community" (the community of Matlab users in this case) is entirely valid and reasonable. In fact, what you and i might think of as "fully open source" is still "reproducible within a community". That community might be the community of people that think it is reasonable/easy to install/use Linunx+Python+Tensorflow.
@drj The point you’re making about "community" is quite interesting. There is indeed an entry cost to making projects reproducible and also to re-execute them. These days, I think that Docker offers a good compromise, but I am also looking into #Nix. In any case, there’s no free lunch...
Add comment