If you delete a program from the Programs folder, does it get uninstalled from the system? Nope. You have to go drag the registry and delete any mentions. You have to go looking for shortcuts and delete all those in multiple locations. If you go to the control panel -> add/remove programs, that works decently, but it’s not guaranteed. It’s been a couple decades since I’ve used windows regularly, but there were all kinds of folders, appdata roaming and that’s just what I remember having to go dig through because there is no real standard for anything. That doesn’t even get into system files and dlls.

There are package managers for Windows too. Choclatey is something I have seen in README.md files for various install instructions.

The Unix-like install locations are for organization. Goes back to Unix-like server oses. Linux is Unix-like as well. Linux doesn’t have a registry to keep track of all the locations for all the files and configurations for each program, so organized locations where the system can expect to find specific things is how it’s done. I much prefer the organized file structure, digging through the registry where there is only the minimal organization was something I always hated.

You can install programs with sudo apt get (program). You can get programs with snaps. You can get programs with flatpaks. You can install with tar.gz files. You can install with .deb files. You can get programs with .sh files.

APT, the Advanced Packaging Tool, uses .deb files. Speaking technically, APT is the app store part of the system, the part that actually installs packages is called dpkg. If you have a .deb file and you want to install it, you can directly invoke dpkg. I have encountered things like printer drivers where the vendor will provide in .deb format. For one reason or another they’re doing it the way the Debian/Ubuntu standard package manager does it, but not publishing the files to the repositories.

Snaps, flatpaks and appimages are intended to solve the problem of fragmentation. See, APT with it’s .deb files are the Debian system used by most if not all Debian/Ubuntu forks. DNF with .rpm files are Red Hat/Fedora’s system. Arch uses Pacman with .pkg.tar files. SuSe, Nix, and a couple others have their own systems, and none of them interoperate.

One thing these do have in common is where they put the files they install; they are sorted by type and function into several directories in the root directory. Most of them you’ll find in the /usr directory, such as /usr/bin, /usr/sbin, /usr/lib or /usr/share. bin for binary files, this is where executables like sh or bash or cat or grep. sbin is for binary files that are only for those with super user privileges; you have to switch users to root or use something like sudo to execute those. Non-executable files like libraries, data, assets etc. will be variously stored in lib, local, share and so forth. You may find there are duplicate folders, like /bin and /usr/bin. In the olden days of minicomputers these were often located on separate drives, but nowadays /bin is a symlink (basically a shortcut) of /usr/bin. Some of that is maintained for backward compatibility or in case you still need to use the differences. It gets a lot more nuanced than I’m talking about here.

They aren’t installed in /home/$USER because if you have multiple users on the machine you probably want them all to have access to the software. Say you’ve got a husband, wife and two kids, would you want to have to install four copies of LibreOffice?

If you do want to install software in the user’s home directory, the place to do that is probably /home/$USER/.local/bin

=====

This does indeed make it a pain in the ass to publish software for Linux. So they wanted one system that would work on most if not all distros. Of course three competitors arrived.

Snap is Canonical/Ubuntu’s attempt which IIRC started with their embedded systems but they’ve been pushing as their whole deal. It’s the closest thing to an Apple App Store there is; it’s owned by Canonical, the back end is proprietary so no one else can host the Snap repository, etc. A lot of folks don’t like that they kept some of it proprietary, plus there are performance issues with how it’s implemented; it mounts virtual disks for each app installed from Snap (or something similar to that) so it gets messy.

Flatpak I think sprung up in or around the Fedora side of the world but has always had the goal of being omnicompatible. It is totally open, you can host your own flatpak repo if you want, though the de facto standard repository is Flathub. At this point, I think Flathub has achieved the goal of being the one place where, if you have a commercial app you want to also publish a Linux version of, publishing it as a Flatpak will reach most of the Linux audience in one shot. There are a couple downsides to it; for example Flatpak isn’t a great way to distribute command line tools, but for GUI-based applications it works fine. Flatpaks are weird; they tend to install in /var somewhere, I think for reasons surrounding permissions.

Appimage is also around; the technique here is it’s basically a .iso file that has everything the app will need to run in it, so you mount the .iso and run the executable file inside. No infrastructure or installer needed. There are package managers for it, but they aren’t required. I find appimage to be the one to use if you’re making some small niche software that 200 people in the world will ever want and you don’t want to bother with the repositories, but I don’t like using them for main distribution.

=====

Let’s change gears and talk about .tar.gz or .run or .sh files. These are non-standard ways of installing software on Linux and I would avoid them unless it’s a last resort type of situation.

A .tar.gz file is functionally similar to a .zip file on Windows; it’s a TAR file (Tape ARchive. A bunch of files packaged into one continuous file, initially designed to be written to magnetic tape) which has been compressed with GZip. When software is distributed this way, it is up to the user to extract and store it…somewhere. Probably the best place to do this is in /opt, which is the standard location for pre-compiled software manually added to the machine by the administrator.

.run files have basically no standard and can do damn near anything and are a bad idea. It could be written in any language and do anything. You just don’t know. The one I encountered the most was Simplify3D’s installer; what they did was directly translate their Windows installer to Linux, the way Windows install wizards work, rather than do things in a Linux way. Hopefully you don’t have to deal with that.

a .sh file is a shell script. Once again these can do anything but it’ll be a text file full of shell commands that you can read and understand. It’ll probably consist of a series of commands like wget or curl which download files from the internet and then commands to put them where they need to go in the file system. Once again this tends to be the approach for the “I’m usually a Windows guy but here’s my Linux version” folks. This is going to choose where to put software for you, hopefully in /usr/local.

=====

You may also find yourself compiling software from source code, which is what’s happening when you’re told to go to Github and then run a series of instructions like git clone, make, make install. Make is basically a scripting system specifically for compiling software, it will figure out what to do. I believe locally compiled software ends up in /usr/local/bin.

===

Summary: There are a lot of directories where software ends up, and it’s done the way it is so you can tell by where it is stored what it is, what it does, who has access to it, who manages it and how it arrived on the system. It also keeps automatic processes like the package manager from interfering with software manually installed or compiled by the users or system administrators.

Why not consolidate all this?

https://xkcd.com/927/

But, there is actually a location standard: https://en.wikipedia.org/wiki/Filesystem_Hierarchy_Standard

This is part of why these days I just stick to flatpaks. No fragmentation, same on any distro, I know where all the programs are going as well as all their config files.

If I want to back up my flatpaks I can do so trivially.

It’s a godsend. Way better than having a bunch of different formats everywhere, or the windows-style some programs installed in XYZ directory, some in program files, some in program files (x86), with config files saved literally anywhere. Maybe it’s in one of the dozens of poorly laid out appdata folders, maybe it’s where the exe is, maybe it’s in documents, maybe it’s in C:, maybe it’s hidden in my user directory, etc. I’ve even seen config files saved to bloody onedrive by default, leading to some funky app behaviour when I wasn’t connected to the internet, or when I ran out of onedrive space.

Linux doesn’t really have stand alone programs like Windows. A package is a series of files that get placed in the proper places plus some optional scripting. Packages have dependencies so you can’t just run a binary from a package. The closest thing Linux as is AppImage but it lost a lot of steam.

In Linux there are two general types of package managers. The first one is native packages. Native packages install to the root filesystem and are part if the core system.

The second type of package manager is the portable format like Flatpak. Flatpaks can either be installed system wide or as a local user. The big difference is that they run in there own environment and have limited permissions. This is done by creating a sandbox that has its own filesystem so that it is independent of the system. This is also what makes them portable as that environment is the same no matter what.

Technically snap packages are portable but you aren’t going to see much use outside of Ubuntu since the underlying architecture has so many flaws.

The idea is to not manage programs by hand to avoid messes.

The multiple solutions come from various needs, but it’s more of an underlying complexity of installation than a real intent to have so many.

Ideally we would need one clean package manager that handles everything without ever having to tinker with installation paths and various methods.

I think that, while, yes, fragmentation hinders a system, it is also its saving grace, as it also stops a given family of systems from growing into what made the competition problematic.

Taking the Program Files folders as example, they have limited read/write permissions on Windows, so whenever possible, I try to install them onto a folder I make in the root of C:. But more and more, since at the very least Windows XP from what I could observe, Microsoft is training users into using only the users folder, and less and less programs give an option to install elsewhere, installing only on the Program Files folder instead. Meanwhile, on Linux Mint (my distro of choice), if AppImage (my to go medium of programs) isn’t working well, I can always fallback to other means, such APT directly or downloading its .deb files then extracting them, getting from flatpak, compiling it myself, building a custom AppImage, running on a VM or emulator, or in the worst possibility, I make a dual boot between Mint and some other distro.

Also, although there are many package managers, from my experience, they usually work similarly. Some changes in syntax, options and names, but nothing outlandish. It would be, I think, like someone learning a close language to his/her mother tongue. And from experience, you can even organize installations in a more standardized way, although it will take some effort from your part to figure out how, since some adaptations may be needed (java 8 and sdl ptsd intensify).

And lastly, from what I can observe, stuff in Linux more often than not share logic or even methods with a lot other stuff in the system. Dunno if it’s a bit of a bias of someone that’s using Linux for a few years already, but the fragmentation usually feels superficial to me, with distros being more tweaks of the ones they stem from, and major changes being better observable when distros are sufficiently far apart.

there’s different ways to install things because they each have their use cases in which they’re better than others (or used to have use cases)

• binary package managers (e.g. apt): fast and lightweight because it only downloads/installs the necessary binaries

• flatpak: can be installed on any distro, but takes up more storage space because they’re installed in a sandbox and all the dependencies are also installed with it, for every application

• snap: same thing as flatpak but a bit worse, but some applications are only packaged for snap because canonical paid a lot of big companies to package for snap (rhey didn’t incentivise against flatpak, they just didn’t fund flatpak)

• appimage: the ‘windows exe’ kinda thing and has all the dependencies bundled so distro agnostic, but you have to manage the appimage files yourself unless you get a manager for it and you can’t update them centrally like you can do with other stuff

• source code repos (e.g. aur): have to compile every new version yourself on your machine, so is slow to update, but often offers things not in the binary package manager

• .sh files for installation: idk why these are used, they’re just annoying. a lot of proprietary software from corpos use them (probably so they can verify dependencies themselves and not trust the system)

• binariy files (e.g. .deb): same thing as with appimage except they’re not distro agnostic

• tar.gz: is just a compressed file format like zip