Scaffolding Installation for Online Workshops

Installing software is hard!

If you have been involved with a Carpentries workshop (or any technical workshop!), you have probably noticed that ensuring learners have a working installation of the tools that will be used can be a significant challenge. Installing software is an acquired skill, and troubleshooting the installation process often requires some of the very skills that we teach in our workshops! Some Instructors pre-empt these issues by offering an “installation party” the evening before the workshop - however, folks who have the least experience with software installation are the least likely to recognise how challenging it can be, and therefore often the least likely to show up for these installation help events.

At in-person workshops, Instructors and Helpers are able to work individually with learners at the start of the workshop to overcome installation problems. This often requires looking over the learner’s shoulder, discussing issues with other Helpers, and (occasionally) going hands-on on the learner’s keyboard to fix complex issues. Online workshops pose their own unique challenges with installation. Although breakout sessions are available on some platforms, it is not easy to sort learners who need help into breakout sessions with appropriate helpers, or for helpers to hop across breakout rooms to provide support to different learners.

The Carpentries teaching philosophy emphasises learners using their own computers, as this leaves them more prepared to continue learning after the workshops. Because of this, we strongly recommend learners install all of the required software on their own computer, however, we also want to support our community in adjusting to the unique challenges of online workshops. We are providing cloud instances for each curriculum, pre-loaded with all required software and data. These are intended to function as backups for learners who are experiencing significant installation difficulties. In the rest of this post, we describe the approach we’re currently taking in providing these backups, some of the design choices that went into developing this solution, and how we are collecting feedback and suggestions about alternative solutions.

What solution are we proposing?

Our primary aim in developing an installation solution is to meet learners where they are, and does not require knowledge of any of the tools they will be learning in their workshop. We anticipate that learners may be frustrated, and strive to offer a user-friendly solution. After exploring different options, we have settled on two main platforms to provide “one-click” installation for learners - RStudio Cloud for R-based workshops, and My Binder for Python-based workshops.

These platforms have strengths and drawbacks, but have both been used extensively for classroom teaching, have active user communities, and are well-supported. In fact, Binder buttons already exist for some of our lessons, and we already use RStudio Cloud in some Data Carpentry workshops.

We have set up a collection of pre-configured Jupyter and RStudio instances on these platforms for all lessons in the official Data Carpentry, Library Carpentry, and Software Carpentry curricula. Because these are intended to serve as backups for learners facing installation difficulties, and we want to continue to encourage learners to install software on their own computers, we will provide the relevant links to Instructors by email in advance of their workshop. Instructors can then share that information with learners who have not completed install by the start of the workshop, allowing them to quickly jump into participatory live coding.

R-based workshops

For R-based workshops, our backup solution uses RStudio Cloud. This has a nearly identical interface to the desktop version of RStudio, has a quick package installation feature, and gives the user the ability to save their work (after logging in with a Google, GitHub, or Rstudio Cloud account). Once the learner clicks the link provided by their Instructor, the RStudio session should start up within 30 seconds to 1 minute. Both Shell and Git lessons can be taught from the “Terminal” tab in RStudio (see caveats below).

Caveats:
The learner must sign in to RStudio Cloud and copy the project to their own account (this is done by clicking a button at the top right hand side of the page).
For shell and git lessons there is no nano editor, so the learners must use the text editor provided within RStudio.

These caveats, and other usage instructions, will be shared with Instructors, and we are currently determining the best “home” for these instructions on The Carpentries website.

Python-based workshops

For Python-based workshops, our backup solution uses MyBinder. This has an identical interface to Jupyter Hub and Jupyter Lab and does not require learners to log in. Once the learner clicks the link provided by their Instructor, the session should start up within 30 seconds to 1 minute, if it has been used within the previous 12 hours. Shell and git lessons can be taught from a new Terminal session within MyBinder, but git lessons including the use of GitHub should not be taught through the MyBinder interface (see caveats below).

Caveats:
As learners are not required to log in, they will need to save and re-upload their files (including notebooks and cleaned data) between sessions. must be saved and re-uploaded between sessions.

The session will stop after 10 minutes of inactivity.
The first time the notebook is used, and If the notebook has not been used within 12 hours, it can take up to 10 minutes to start up.
Learners should NOT connect to GitHub using MyBinder. All information entered into the console should be treated as public (SOLUTION: use RStudio Cloud for Git lessons [see below])

Shell lessons

For a pure shell workshop (no R, no Python, no Git), our backup solution uses MyBinder. This has an interactive shell, manual pages, and the nano text editor installed.

Caveats:
As learners are not required to log in, they will need to save and re-upload their files (including notebooks and cleaned data) between sessions. must be saved and re-uploaded between sessions.

Git lessons

For any workshop where GitHub will be used, our solution is to use the RStudio Cloud terminal tab. This is the only online interface that will be secure enough to allow people to connect with GitHub.

Caveats:
RStudio Cloud does not have the nano text editor installed, so the learners must use the text editor provided within RStudio.

These caveats, and other usage instructions, will be shared with Instructors, and can be found in the Scaffold Support Guide.

How will this information be shared with Instructors and learners?

Instructors will receive links to these backup solutions in one of the pre-workshop preparation emails sent by our workshop coordinators. This email will also include information about the caveats discussed above, and a link to detailed instructions for learners, including screenshots. We recommend that Instructors wait to share this information with learners until the start of the workshop, and then only with those learners who have not successfully completed installation on their own computer.

What other options have we considered?

Examples of services that looked promising, but were ultimately rejected were:

Google CoLab: This is an interface to Jupyter notebooks that’s hosted by Google and stylized to blend with Google Drive. It was rejected because it required a Google account, usage limits were unclear, and the interface was sufficiently different from Jupyter Lab
CoCalc (from Sage): This interface has nearly all of the tools needed, but was ultimately rejected because it has a sufficiently different interface and sets defeating expectations for users on the free tier (e.g. “expect poor performance… “, “the free servers are randomly rebooted frequently,…” [emphasis theirs])
Cyverse Atmosphere: This service is largely oriented towards bioinformatics users, but has a lengthy registration form and requires a payment plan to use the cloud computing services

We will be collecting feedback from instructors about how many of their learners needed to use these backup options, so that we can scale up the infrastructure if needed. We are also reaching out to individuals in our community who have experience working with MyBinder or RStudio Server to ask for their guidance. Last, but certainly not least, we are seeking input from a small group of recently-certified instructors who have not used these tools in the past. This last set of reviewers will help us overcome our expert awareness gap and ensure the solutions we’re providing are usable by novice learners. We encourage feedback from any of our community members. We recognise that every workshop will be different, and encourage experienced instructors to test out other installation solutions and report back how things work! Please provide us your feedback either by opening an issue in the Scaffolds repository or by email at team@carpentries.org.