Linux From Scratch for Computer Science final project

Posted by Astronaut_Striking@reddit | linux | View on Reddit | 11 comments

Hi all, I'm brainstorming ideas for my final project and I'm wondering about the feasability of creating a linux distro from scratch, for this project. I want to learn the ins and outs of Linux systems and this seems to be the best way of doing so, and if I create a bootable ISO from it, there will be a final 'product' that people can use and I can show off in a presentation. Just wondering if anyone's done something similar to this or if anyone can chime in on whether or not it could be a real idea for a final project. If it's something I could really stick on a portfolio or resume and benefit from. I know it is something I want to do, I just haven't really heard anyone mention it as a project before. Going to mention it to my project supervisor so I can get an official opinion, but thought I'd see what the general opinion of it is on here :)

Reply to Post

11 Comments

[-]

RusselsTeap0t@reddit

Is this "trolling"? Building Linux from Scratch is very easy and there is a project for it that's been around for decades. You just copy and paste commands but it can be time consuming based on your hardware: [https://www.linuxfromscratch.org/lfs/view/stable/](https://www.linuxfromscratch.org/lfs/view/stable/) If you search for a more sophisticated version here is [another project](https://github.com/dslm4515/CMLFS): It uses: \- **Clang** instead of GCC as a compiler. \- **Musl** instead of Glibc as a C library. \- **Elftoolchain** instead of Binutils as binary utilities. \- **S6** instead of Sysvinit for the init system. \- **Busybox** instead of GNU Coreutils. \- **LibreSSL** instead of OpenSSL. \- **Gettext-tiny** instead of Gettext.

[-]

Astronaut_Striking@reddit (OP)

Not trolling, just asking. I didn't know it would be a very easy and quick project to complete. The idea is that I create a whole system from scratch, learning about all of the components of said system and having an operating system I can install on machines.

[-]

RusselsTeap0t@reddit

There is already a thing called "Linux From Scratch" for decades. It's actually a free book, you can find it as PDF or you can also read it directly from the website. I sent you the link. It explains every single part of a Linux system and how you can build it from scratch. At first you use a normal Linux system (distro does not matter) to build a cross-compiler. After building the cross compiler, you build some needed tools for the target system until you can chroot the system. Then you chroot into the system to build other stages and needed parts. At the end you will have a very minimal Linux system with its base. You can of course use it or reproduce it as an Operating System.

[-]

Astronaut_Striking@reddit (OP)

Thank you! I was referring to Linux From Scratch or anything similar to that. In terms of scale, is this something you could spend 6+ months working on or is it a much smaller task that I’ve completely overestimated? 😂

[-]

RusselsTeap0t@reddit

3 hours max with i9 9900k or newer CPUs for compiling. And your reading speed for reading the whole book to learn. If you want to write the book yourself without help it would take 6 months to research, try and debug.

[-]

Astronaut_Striking@reddit (OP)

Ahh okay. The whole project would include obviously building the system, designing it to be something a user could stick on a flash drive and install with with a GUI like Calamares and feel as fleshed out as possible. A lot of the work is also documentation, creating a prototype, meetings about progress, etc. It wouldn't just be a case of building the system and calling it a day, I'd have to show a true understanding of every step as at the end we present the project to a small group that will essentially interrogate you to make sure it really is your own work, and that it does serve a real purpose.

[-]

RusselsTeap0t@reddit

Even then it shouldn't take too much time like months. Because there are already distros that use LFS as a base. It's not so niche I mean. The biggest problem you will have will be the package management. There are some solutions though but you need to read the book well as well as some extra resources. 1. You read the book. 2. You build the system. 3. You implement some kind of a package management system. 4. You back up the system to reproduce. 5. You create documentation. Keep in mind that there is another (much bigger) book for you to look at that is BLFS (Beyond Linux From Scratch). It also explains the details after you build an LFS system.

[-]

RusselsTeap0t@reddit

I mean easy for building it, not understanding. If you want to understand it completely you need to read thoroughly along with extra material.

[-]

mauros_lykos@reddit

> ideas for my final project and I'm wondering about the feasability of creating a linux distro from scratch Is such a project acceptable as your final thesis?

[-]

Astronaut_Striking@reddit (OP)

The project aims to product a technical artefact such as a piece of software, this would fit that critera. However, the project should aim to be the solution to a problem. It's unclear what the scale of that problem needs to be, as some recommended ideas are things such as an FPS game and a VR game. It could be the case that for those the problem is boredom, which is solved by a video game. I'm sure a linux distro could be aimed to solve a problem, giving purpose to a machine which otherwise has no use (without an OS), allows for users to access the internet, complete work, etc. I'm going to ask about it with a supervisor, once we've either selected one or been allocated one. I want to be creative, but I don't want to over complicate it and create something which has gone far off the rails of what the project is supposed to accomplish.

[-]

mauros_lykos@reddit

>I'm going to ask about it with a supervisor, Start from that. I believe they won't accept it.