Offline Capable Linux Lab

We need an offline capable Linux teaching lab.

Constant access to the Web is a double edged sword. There are clear benefits, but for a beginner practicing programming or any similar set of cognative skills, the option to respond to any challenge by turning to the web can allow them to progress through learning material without really engaging with it in a way that teaches them the skills they need to learn.

Nothing in an introductory programming sequence requires internet access - a stand-alone workstation is sufficient. All of the tools and documentation can be installed locally. The concepts and techniques can be covered in a paper textbook, or an e-book on disconnected workstation.

Teaching some of our early programming courses (e.g. Intro to Programming, Data Structures, System Programming in C) in an offline environment would allow the students to learn the material better and be better prepared to use the Web to enhance their personal skills. These skills will be especially valuable for them in later courses, where trying to find solutions rather than personally solving problems is much less feasible.

This offline lab should run modern desktop Linux. Setting up an offline environment will be easier on Linux. We don’t currently have a Linux lab, which creates multiple holes in our program. And taking advantage of some of the more interesting opportunties of a dedicated offline-capable lab would be easier on the Linux platform.

Offline Lab Assignments

#

Do exercises 10.6, 10.8, and 10.9 in the textbook.

In our current labs, that sort of assignment doesn’t work. Best case, the students type the questions into Google. Worst case, they find an answer sheet or go to an AI chatbot.

With an offline lab, that sort of assignment would become possible. Further, standard lab assignments work better too and become more reusable.

To get the benefit, lab assignments would need to be separate from homework. But this should even help longer homework assignments, by giving the students a clear idea of what the assignments are practicing for.

Offline Programming Exams

#

Write a function that uses recursion to find the longest string in a list of strings. Your function must be correct and must pass the provided test cases.

It’s currently possible to give questions like that on paper exams, but the fact that the student can’t compile or run the code nor access the standard documentation makes such questions much slower and clunkier than they need to be. On paper there can’t be as many questions, nor can they be as detailed, nor is exactly the right skill being tested.

With an offline lab, it would be possible to to have students write a real program using standard tools and go through the typical code / test cycle.

Mechanics

#

The simplest version of this would be:

• A lab full of desktop computers
• Desktop Linux installed normally on each computer
• The lab configured as a stand-alone LAN.
• A simple gateway/router would issue local IP addresses and provide local DNS name lookup.
• A single ethernet uplink specifically intended to be easily physically unplugged.

There are some compliations:

• Each student should have their own user account, which works on any workstation, and allows for authentication when the lab is offline.
• It would be nice for each student to have a single home directory, accessible from any workstation.

Both of those can be handled by having a single in-lab server providing LDAP for authentication and NFS for home directories. Providing login with USNH credentials is worth exploring, but not worth any compromises to local functionality - creating local accounts is easy.

Once we have an in-lab storage server for NFS home directories, it should be straightforward to allow the workstations to network-boot. This would let us:

• Have multiple images to boot from, configured with different operating systems for diffrent classes and use cases. For example, the image for an exam might be different from the image for a lab session.
• This would mean every boot would get a clean OS image, which has some advantages.
• We could even let students have root access on workstations with either by not mounting home directories in that case or by putting in a little bit of effort at setting up LDAP and Kerberos to authenticate NFS mounts.
• With a local distro package mirror, we could allow packages to be added to boot images or to be installed using local root even when the lab was offline.

It is explicitly a goal of this project to not require access to or interoperation with any system-wide USNH cloud infrastructure.

The disconnect mechanism should be a visually obvious physical disconnection - preferably a big Dr. Frankenstein style switch - not some software thing and certainly not anything that maintains a network connection for limited cloud connectivity or remote management.

Submitting Assignments

#

There are two main ways students could submit lab assignments completed in this lab:

• Connect the lab to the internet a few minutes before the end of the period and have the students submit to Canvas or Inkfish.
• Have the students submit their work to a service on the local server, either through a web browser or a command line script.

Either way students would not have access to the web for the majority of their task time, and would submit their work to a central location that the instructor can access for evaluation and feedback.

Secondary Benefits

#

Although my primary core is to support introductory programming courses, this resource would also be useful for projects in later courses.

Further, this would provide the classic benefit of exposing students to examples of a range of network and computer concepts being applied in practice. For example, it would show that computers and even a network of computers will still work even when not connected to the internet.

Cost

#

This should easily be able to use existing machines from within USNH, including for the server and gateway. The optimal hardware would be reasonably low spec and not brand new for maximum software compatibility - this would be an excellent use for a bunch of machines that were otherwise end-of-life.

The main concern would be how this overlaps with existing lab resources. Designing it to allow the machines to run managed Windows and and therefore be used for non-CS classes would drastically increase the complexity, probably to the point of infeasibility.

I think the baseline question here might be to ask why we can’t simply convert 419. It already has a really obvious rack for the main network drop.