RDM Monday
Summary: In our bioinformatics research a significant part of our experimental setup is tangled up in computers and the software they run. Even though computers and software are fully explainable and potentially deterministic, we usually find ourselves lacking the tools to describe this computational realm comprehensively and in unambiguous terms. This prevents our peers from reproducing, validating, and building upon our experiments. In this session we will look at the problem of computational reproducibility, demystify containers, and introduce the declarative approach to reproducible and portable software environments using Guix [giks]. Topics covered: - why care about software reproducibility? - reproducibility is trickier than you may think: why version strings aren't enough. - stateful vs declarative environments: finding the balance between restrictive planning and free experimentation - using "guix shell": environments and containers. - using "guix time-machine" - manifests: executable declarations of reproducible state - compromises: using Guix as a pragmatic reproducible foundation Keywords: guix, reproducibile computational science, containers, software methodology for publications, best practices, pragmatic software management |
Venue
Virtual Meeting