tmcphillips / openrefine-provenance Goto Github PK

Binder provides several means of configuring a Docker container for working with the code and data in a repository. The postBuild step allows a custom Bash script to customize the Docker container. Try running Ansible playbooks in postBuild to recreate the computing environment for this repository so that it can be used through Binder.

The bootstrap.sh script I have been using to configure a WSL environment from scratch assumes that I want to tie the WSL user account to the Windows account of the same name, share .ssh directories with it, etc.

Add a flag for enabling this tight integration so that by default the bootstrap script creates a relatively isolated environment. The bootstrap scripts for Vagrant-managed VMs can do the same thing, and and share code with the WSL Debian bootstrap script.

The bootstrap scripts and ansible playbooks currently are versioned in tmcphillips/ansible-playbooks.

I have been using a Debian environment installed in the Windows Subsystem for Linux for the initial investigation of OpenRefine. This lets me do the work in a Linux shell, with software installed via Ansible, and at the same time use Windows desktop applications to edit files and view results.

Include the Ansible playbooks used to configure this environment in this repo.

Whole Tale, Binder, and the commercial platform Code Ocean employ Docker containers to create and maintain computing environments for reproducing scientific computations. However, I find working within Docker containers to be limiting when doing research in the first place. My guess is that many folks who share their completed research via Docker actually do the real work of research outside of Docker containers, and use Docker to emulate their real work environment for sharing with the community at the end of the project. I would prefer to work in a computing environment that is reproducible in the first place, and to share that environment with the community.

List the limitations of Docker for creating reproducible computing environments for day-to-day work so that alternative approaches can be evaluated.

Vagrant is cross-platform and makes it easy to configure and manage virtual machines on Linux, MacOS, and Windows. If the research computing environment can be recreated using Vagrant and Ansible, then it can be reproduced on all three platforms.