Yeah I was thinking the same. Perhaps make a sticky post about it once a week.

A symlink works more closely to the first way you described it. The software opening a symlink has to actually follow it. It’s possible for a software to not follow the symlink (either intentionally or not).

So your sync software has to actually be able to follow symlinks. I’m not familiar with how gdrive and similar solutions work, but I know this is possible with something like rsync

Software opens a symlink the same way as a regular file. The kernel reads a path stored in a symlink and then opens a file with that path (or returns a error if unable to do this for some reason). But if a program needs to perform specific actions on symlinks, it is able to check the file type and resolve symlink path.

To determine how some specific software handle symlinks, read its documentation. It may have settigs like “follow symlinks” or “don’t follow symlinks”.

Whenever I open the symlink, does the software (player) understand «oh this file seems like a symlink, I should go and open the original file», or it’s a filesystem level stuff and software (player) basically has no idea if a file I’m opening is a symlink or the original movie.mp4?

Others have answered well already, I just will say that symlinks work at the filesystem level, but the operating system is specially programmed to work with them. When a program asks the operating system to open a file at a given path, the OS will automatically “reference” the link, meaning it will detect a symlink and jump to the place where the symlink is pointing.

A program may choose to inspect whether a file is a symlink or not. By default, when a program opens a file, it simply allows the operating system to reference the file path for it.

But some apps that work on directories and files together (like “find”, “tar”, “zip”, or “git”) do need to worry about symlinks, and will check if a path is a symlink before deciding whether to reference it. For example, you can ask the “find” command to list only symlinks without referencing them: find -type l

Symlinks are fully transparent for all software just opening the file etc.

If the software really cares about this (like file managers) they can simply ask the Linux kernel for additional information, like what type of file it is.

its a pointer.

E: Okay so someone downvoted “it’s a pointer”. Here goes. both hard links and symbolic links are pointers.

The hard link is a pointer to a spot on the block device, whereas the symbolic link is a pointer to the location in the filesystems list of shit.

That location in the filesystems list of shit is also a pointer.

So like if you have /var/2girls1cup.mov, and you click it, the os looks in the file system and sees that /var/2girls1cup.mov means 0x123456EF and it looks there to start reading data.

If you make a symlink to /var/2girls1cup.mov in /bin called “ls” then when you type “ls”, the os looks at the file in /bin/ls, sees that it points to /var/2girls1cup.mov, looks in the file system and sees that it’s at 0x123456EF and starts reading data there.

If you made a hard link in /bin called ls it would be a pointer to the location on the block device, 0x123456EF. You’d type “ls” and the os would look in the file system for /bin/ls, see that /bin/ls means 0x123456EF and start reading data from there.

Okay but who fucking cares? This is stupid!

If you made /bin/ls into /var/2girls1cup.mov with a symlink then you could use normal tools to work with it, looking at where it points, it’s attributes etc and like delete just the link or fully follow (dereference) the link and delete all the links in the chain including the last one which is the filesystems pointer to 0x123456EF called /var/2girls1cup.mov in our example.

If you made /bin/ls into a hardlink to 0x123456EF, then when you did stuff to it the os wouldn’t know it’s also called /var/2girls1cup.mov and when /bin/ls didn’t work as expected you’d have to diff the output of mediainfo on both files to see that it’s the same thing and then look where on the hard drive /var/2girls1cup.mov and /bin/ls point to and compare em to see oh, someone replaced my ls with a shock video using a hard link.

When you delete the /bin/ls hardlink, the os deletes the entry in the file system pointing to 0x123456EF and you are able to put normal /bin/ls back again. Deleting the hard link wouldn’t actually remove the data that comprises that file off the drive because “deleting” a “file” is just removing the file systems record that there’s something there to be aware of.

If instead of deleting the /bin/ls hardlink, you opened it up and replaced the video portion of its data with the music video to never gonna give you up, then when someone tried to open /var/2girls1cup.mov they’d instead see that music video.

if that is, the file wasn’t moved to another place on the block device when you changed it. Never gonna give you up has a much longer running time than 2girls1cup and without significant compression the os is gonna end up putting /bin/ls in a different place in the block device that can accommodate the longer data stream. If the os does that when you get done modifying your 2girls1cup /bin/ls into rickroll then /bin/ls will point to 0x654321EF or something and only you will experience astleys dulcet tones when you use ls, the old 0x123456EF location will still contain the data that /var/2girls1cup.mov is meant to point to and you will have played yourself.

Okay with all that said: how does the os know what to do when one of its standard utilities encounters a symlink? They have a standard behavior! It’s usually to “follow” (dereference) the link. What the fuck good would a symbolic link be if it didn’t get treated normally? Sometimes though, like with “ls” or “rm” you might want to see more information or just delete the link. In those cases you gotta look at how the software you’re trying to use treats links.

Or you can just make some directories and files with touch and try what you wanna do and see what happens, that’s what I do.

Because Linux and the programs themselves expect specific files to be placed in specific places, rather than bunch of files in a single program directory like you have in Windows or (hidden) MacOS.

If you compile programs yourself you can choose to put things in different places. Some software is also built to be more self contained, like the Linux binaries of Firefox.

Expanding on the other explanations. On Windows, it’s fairly common for applications to come with a copy of everything they use in the form of DLL files, and you end up with many copies of various versions of those.

On Linux, the package manager manages all of that. So if say, an app needs GTK, then the package manager makes sure GTK is also installed. And since your distribution’s package manager manages everything and mostly all from source code, you get a version of the app specifically compiled for that version of GTK the distribution provides.

So if we were to do it kind of the Windows way, it would very, very quickly become a mess because it’s not just one big self contained package you drop in C:\Program Files. Linux follows the FSH which roughly defines where things should be. Binaries go to /usr/bin, libraries to /usr/lib, shared files go to /usr/shared. A bunch of those locations are somewhat special, for example .desktop files in /usr/share/applications show up in the menu to launch them. That said Linux does have a location for big standalone packages: that’s usually /opt.

There’s advantages and inconveniences with both methods. The Linux way has the advantage of being able to update libraries for all apps at once, and reduce clutter and things are generally more organized. You can guess where an icon file will be located most of the time because they all go to the same place, usually with a naming convention as well.

different strokes.

windows comes from the personal computing world and retains a bunch of stuff from it to this very day for no good reason, in this case there used to be no guarantee that a particular installation target would have the target directory mapped in a consistent way so the installer would make a guess and give the user a chance to change it.

if that sounds stupid, it is. no one writes in assembly anymore, they target the OS and nowadays the OS will have a consistent set of folders to install stuff to. we all know where the program “should” be installed to already.

but it didn’t used to be like that in the PC world! used to be your computer wasn’t a fixed purpose windows computer from the jump, never to be anything else. there were different OSes that people would use regularly and even different DOS environments which a person could use to run programs under. Hard disks weren’t disks inside the machine, but big beige external disks that you’d plug up, set beside the computer and access after booting. in that setup where a programmer targeted DOS (if they cared about the execution environment at all and didn’t just write for the processor) it made sense to ask where someone was gonna want to install their software, and to what extent they’d even want to start dirtying up the media they paid good money for with some knuckleheads weird files from some goofy program on a stack of floppy disks.

linux comes from the unix world, where the question of where something installs is easy and straightforward: it installs in $PATH. what is $PATH? it’s where the os will look when you try to run something to see if it can run any program by that name. if a program isn’t installed in $PATH then when you type its’ name in and hit enter the computer won’t know what the hell youre talking about and you’ll have to type it’s whole ass location out and hit enter.

Why didn’t unix systems that linux imitates ask you where to install stuff? because usually it wasn’t your choice! linux was unix for personal computers and unix was run on systems that took up whole rooms with all sorts of equipment. you might be the user of that system but never have access to the room with all the spinning disks and flashing lights, stuck on a terminal dialing in over a serial line.

so the assumption was that you’d have a variable in your user environment that would say where things were installed but not that you’d have the ability to change it or even install things.

so why in a linux environment would you ever install anything outside of $PATH or even want to be sure where something’s installed at all?

even under linux it can be useful to do either. installing outside of path keeps programs from being accidentally autocompleted or invoked. installing in a particular component of $PATH ($PATH can be many directories!) lets you put serious business programs that demand maximum performance on faster media.

so why the hell won’t linux systems give you the option of installing in a specific location or outside of $PATH altogether?

they will, but unlike windows, they don’t ask you. unless you specifically ask to do that unique and very abnormal operation, they just do the usual thing. when you want to install weirdly you gotta dig into your package manager and packaging system. sometimes you unzip a package and change a line in a file then zip it back up and install from your modified version.

Don’t think there is.

system32 holds files that are in various places in Linux, because Windows often puts libraries with binaries and Linux shares them.

The bash in /bin depends on libraries in /lib for example.

What is system32? Outdated 32bit binaries?

Instead of installing packages through a package manager one at a time and configuring your system by digging into individual config files, NixOS has you write a single config file with all your settings and programs declared. This lets you more easily configure your system and have a completely reproducible system by just copying your nix files to another nixos machine and rebuilding.

It’s also an immutable distribution, so the base system files are only modified when rebuilding the whole system from your config, but during runtime it’s read only for security and stability.