Ten Years of Guix — Program
Ten Years of Guix — 16–18 September 2022, Paris, France Program Venue Photos Sponsors The program spanned three days, from Friday to Sunday, 9AM to 6PM local time (CEST) . Below you will find links to supporting material and video recordings of the talks. Friday 16th: Reproducible deployment for reproducible research Friday is dedicated to reproducible software deployment as a foundation for reproducible research workflows , with experience reports from scientists and practitioners, using Guix and other tools.
09:00–09:30 👋 Welcome! 09:30–10:00 Reproducibility of bioinformatics pipelines Sarah Cohen-Boulakia (Université Paris-Saclay, LISN) 📁 📺 Read more... Impossibility to redo a given analysis that run on the same computer a few months ago, failing to re-execute the data analysis described in a peer’s recent paper: recently, we have all experienced computational reproducibility issues. The bioinformatics community had to face with the reproducibility crisis and it has been pioneer in the design of elements of solutions to better reproduce and reuse bioinformatics pipelines.
In this talk, we will introduce and discuss such elements of solutions and present lessons learnt in terms of good practices to follow when analyzing large scientific datasets. 10:00–10:30 Maneage: a proof-of-concept for rigorous reproducible research paper criteria Boud Roukema (Institute of Astronomy, Nicolaus Copernicus University) 📁 📺 Read more... Maneage is a proof-of-concept implementation of a set of eight criteria for long-term reproducible scientific research papers.
The criteria are: completeness; modularity; minimal complexity; scalability; verifiable inputs and outputs; recorded history; linking narrative to analysis; and free and open-source software. The criteria will be briefly presented. An outline of Maneage with its structure primarily based on bash, make, Git and LaTeX, and practical experience using and developing it for peer-reviewed papers, will be presented. 10:30–11:00 ☕ break 11:00–11:30 Helping research on distributed systems with a functional package manager Olivier Richard (Inria Grenoble – Rhône-Alpes, LIG) 📁 📺 Read more...
Development of environments for distributed systems is a tedious and time-consuming iterative process. The reproducibility of such environments is a crucial factor for rigorous scientific contributions. Based on the Nix functional package manager we propose a tool that generates reproducible distributed environment. Moreover, it enables users to deploy their environments on virtualized (Docker, QEMU) or physical (Grid’5000) platforms with the same unique description of the environment.
After the presentation of the tool and its benefits, limitations and lessons learned we will be discussed. 11:30–12:00 Guix as a tool for computational science Konrad Hinsen (Centre de Biophysique Moléculaire, CNRS Orléans; Synchrotron SOLEIL, Saint Aubin) 📁 📺 Read more... Guix is described as a package manager and a GNU/Linux distribution. While technically correct, this summary hides the fact that Guix is a valuable support tool for reproducible computational research, as an add-on to any GNU/Linux distribution, much like many of us currently use Docker containers.
I will attempt to convince you that Guix is a better choice than Docker, and show you how to get started with integrating Guix into your computational work environment. 12:00–12:30 Software Heritage to the rescue of reproducible science Antoine R. Dumont, Valentin Lorentz (Software Heritage) 📁 Read more... This talk is not recorded but you can read its transcript . Software Heritage is the universal source code archive.
It enables reproducible research and build tools (such as Guix and Nix) by storing source code forever with a public API, which can be used as a fallback when original software artifacts are deleted or unavailable. In this talk, we will discuss how Software Heritage performs global deduplication using a Merkle DAG, and some planned developments, such as reconstructing original tarballs exactly using Guix’s Disarchive .
12:30–14:00 🍽 lunch 14:00–14:30 Archive, reference, describe and cite software source code: a pathway to reproducibility Morane Gruenpeter (Software Heritage, Inria) 📁 📺 Read more... Software is a fundamental pillar of modern scientific research, and it is essential to properly archive and reference the source code of software used in all research activities. In this presentation, we will show how to use the Software Heritage framework to transparently archive all software source code and how to obtain and use the intrinsic identifiers, specially designed for software source code, which are necessary to reference the source code in a way that facilitates long-term reproducibility.
We will also show how to describe the source code for a better understanding of the software and to allow discovery of the software on search engines. Finally, we will discuss the challenges when it comes to citing software to give credit to authors in the academic ecosystem. 14:30–15:00 Using Guix for scientific, reproducible, and publishable experiments Philippe Swartvagher (Inria) 📁 📺 Read more... Ensuring experiment reproducibility with complex software stack can be quite a headache.
This presentation will tell the story of a PhD student discovering Guix to ease the reproducibility of the software environments for scientific experiments. We will cover the motivation to use Guix, the features that help to easily customize software environments to fit the experiment requirements, and then how to share experiment scripts in scientific publications to be able to reproduce the experiments. 15:00–15:30 Camlboot: debootstrapping the OCaml compiler Nathanaëlle Courant (OCamlPro, Inria) 📁 📺 Read more...
In this talk, I will present Camlboot , a project which debootstraps the OCaml compiler, that is, is able to compile the OCaml compiler without using the bootstrap binary. Camlboot consists in a naïve compiler for a subset of OCaml called MiniML, and an interpreter for OCaml written in MiniML. I will first justify the interest of debootstrapping, then explain the architecture and parts of Camlboot, and finally present the experimental validation of Camlboot.
15:30–16:00 GNU Guix and the RISC-V future Pjotr Prins (University of Tennessee Health Science Center) 📁 📺 Read more... RISC-V is a modern open hardware platform that is not only part of many new devices, but also powering up new areas of research that will lead to optimised hardware solutions for offloading tasks to dedicated modules on systems-on-chip (SOCs). The European Union, the National Science Foundation (NSF) in the US, and NLnet foundation recognise this promise and fund the design and development of new solutions, including for high-performance computing (HPC).
GNU Guix, because of its hackability, flexibility, determinismistic output, and potential for ‘generics’, makes an ideal partner for RISC-V development and deployment. In principal the full software stack with all dependencies can be ‘carved in stone’ and provide a reproducible design of RISC-V hardware all the way from idea to taping out a chip, via the stages of simulation, emulation and testing. In this presentation I'll talk about this future of open hardware architecture that allows for ‘burning software’ into hardware and how GNU Guix can play a central role in a new industry.
16:10–17:10 ☇ lightning talks 16:10–16:20 Building a library: from Autotools to Guix Evgeny Posenitskiy (LCPQ, CNRS) 📁 📺 16:20–16:30 Guix quirks when packaging a C GNU Autotools program distributing also a Python package Philippe Swartvagher (Inria) 📁 📺 16:30–16:40 Taming the Python Alice Brenon (ICAR [UMR 5191], LIRIS [UMR 5205]) 📁 📺 16:40–16:50 Demo: a malleable GUI for Guix Konrad Hinsen (Centre de Biophysique Moléculaire, CNRS Orléans; Sy
The rest of this page is on the live site.