The range of Linux is enormous. It is everything from small microcontroller-ish devices to cars, routers, phones, appliances, and servers. These are the primary use cases. Desktop is a side thing.

Part of the learning curve is that no one knows the whole thing top to bottom end to end at all levels of source. Many entire careers and PhDs and entire companies exist here. You will never fully understand the thing, but that is okay, you do not need to understand it like this.

The main things are that every distro has a purpose. Every distro can be shaped into what you want.

Fundamentally, Linux as the kernel is a high level set of command line tools on top of the hardware drivers required to run on most hardware. The Linux kernel is structured so that the hardware drivers, called modules, are built into the kernel already. There is actually a configuration menu for building the kernel where you select only the modules you need for you’re hardware and it builds only what you need automatically based upon your selection. This is well explained in Gentoo in tutorial form.

Gentoo is the true realm of the masters. It has tutorial documentation, but is written for people with an advanced understanding and infinite capacity to learn. The reason Gentoo is special is the Portage terminal package manager. Gentoo is made to compile the packages for your system from source and with any configuration or source code changes you would like to make. This is super complicated in practice, but if you have very specific needs or goals, Gentoo is the place to go. Arch is basically Gentoo, but in binary form for people too lazy or incapable of managing Gentoo, but where they either already have a CS degree level understanding of operating systems or they are the unwitting testers of why rsync works so well for backing up and reloading systems. It is the only place you will likely need and use backups regularly. The other thing about arch is that the wiki is a great encyclopedia of almost everything. It is only an encyclopedia. It is not tutorial or ever intended as such. Never use arch as a distro to learn on. It is possible, but you’re climbing up hill backwards when far easier tutorial paths exist.

Godmode is LFS, aka Linux From Scratch. It is a massive tutorial about building everything yourself. No package maintainers for you.

Redhat is the main distro for server stuff. It is paid. The main thing it offers is zero down time kernel updates. You never need to reboot. It transitions packages in real time. Most of the actual kernel development outside of hardware peripheral driver support happens at Redhat. Fedora is upstream of Redhat. They are not directly related, but many Fedora devs are Redhat employees. Fedora informally functions kinda like a beta test bed for Redhat. Most of the Redhat tools are present or available in Fedora. This is why the goto IT career path is through Fedora using The Linux Bible. So if you want to run server type stuff or use advanced IT tools, maybe try Fedora.

Here is the thing, you do not need to use these distros. They likely are of no interest to you. All of this bla bla bla is for this simple point, distros are not branding or team sports. They are simply pathways and configurations that best handle certain use cases. The reason you need to understand the specific use case is because these are like chapters of Linux documentation. How do I configure, schedule and automatic some package? Gentoo probably has a tutorial I will find useful. How do I figure out the stuff going on prior to init? LFS will walk me through it. What is init? Arch wiki will tell me.

On the other hand, there is certain stuff to know like how Debian is for hardware modules development, and mostly unrelated to the latter, building one off custom server tools. When you see Debian like on some single board computer where no other distro is listed, that means it probably isn’t worth buying or messing with. It means the hardware is likely on an orphaned kernel that will never have mainline kernel support so it won’t be safe on the internet for long.

That’s another thing. Most of what is relevant is keeping a system safe to be online, meaning server stuff.

OpenWRT is the goto Linux for routers and embedded hardware. You can easily fit the whole thing in well under 32 megabytes of flash. It is a pain in the ass for even a typically advanced Linux terminal user, but that is Linux with a GUI too. The toolset is hard, and has little built in documentation by default.

With very early early 1970’s+ personal computers, crashing and resetting computers was a thing. Code just ran directly on the memory. The kernel is about solving the issue of code crashing everything. The kernel creates the abstractions that separate the actual hardware registers and memory from the the user space tools and software so that your code bug does not crash everything. It is a basic set of high level user space commands and structures to manage a file tree, open, edit, and run stuff. In kernel space, it is the scheduler and process management that swaps out what is running when and where for both the kernel processes and separate user processes. The kernel is not the window manager, desktop, or most of the actual software you want to run.

The other non intuitive issue many people have is sandboxing and dependencies. Not all software is maintained equally. When some software has conflicting dependencies with other software, major problems arise. How you interact with this issue is what really matters and one size does not fit all or even most. This issue is the reason the many distros actually exist. Sandboxing, in almost every context you will encounter, is about creating an independent layer location for a special package’s dependencies to exist without conflicts on your base host system. It is not about clutter management or security, just package dependencies. That is the main thing that each distro’s maintainers are handling. The packages native in the distro already have their dependencies managed for you; they should just work. Maybe you want to use more specific or unrelated things. Well then you need to manage them. Nix is designed especially for this in applications where you need to send your configuration sandbox to other user. Alternatively, people use an immutable base like silverblue and run all non native software from sandboxed dependency containers.