Reminder that there are people on this board (and elsewhere) who still, in (1- 2020), unironically defend the Unix Filesystem Hierarchy.
The hierarchy, like most other things in Unix, was hardly "designed" at all. It merely originated out of historical technical reason.
Any attempt to "describe" it, such as man hier is just a rationalization after the fact: lists.busybox.net
When will autists stop putting Unix on a pedestal and recognize we can do better?
Other urls found in this thread:
i don't see you making a goddamn attempt to do better. at least in eunuchs there is some logical design to everything.
Unix is highly modular, freely available, easy for developers to learn, and compatible with pretty much everything after recompiling. I don't really see the point in inventing a new standard and fragmenting everything.
>>When will autists stop putting Unix on a pedestal and recognize we can do better?
it's not "we can't do better", we most certainly could. Its a matter of "doing better wouldn't be enough of a benefit to be worth the bother and disruption of changing existing software and practices"
Gobolinux
Unix and by extension Linux have a lot of momentum behind them preventing any large scale changes to the base filesystem.
It's more likely that any radically different designs are going to come from future microkernel based systems.
The problem is that people now have to deal with working more in this economy and then cope with how fucked everything's going.
>When will autists stop putting Unix on a pedestal and recognize we can do better?
Start.
I don't see a better alternative.
>doing better wouldn't be enough of a benefit to be worth the bother and disruption of changing existing software and practices
You mean aside from transforming unixoid systems from obscure grey boxes into the operating structure of the masses?
I like job security as much as the next guy, but if I could I would make sure that the people coming after me won't need to learn the arcane magics of Unix.
We need a better file system structure, better command names, a definitive class of "system" commands (i.e. ifconfig should have been updated, not replaced by the new hotness) that get developed together with the kernel, a proper pipeline format (parsing text is a good fallback but an asinine default mode), a complete purge of GNU, an update of the underlying tty system (why the fuck isn't each monitor on a multi-monitor setup natively mapped to its own tty, utilizing the already existent ctrl+FX hotkeys for switching between them? Why haven't we taken a good hard look at control sequences and purged the retarded ones/introduced new useful ones for the current year), a standardized config workflow (standard config syntax, as well as a standard format for describing config options, which would double as documentation for all config options a tool has to offer, as well as a standard location for configs instead of our current semi-standardized locations), a true replacement for X (how about having a type of X-forwarding that doesn't guzzle bandwidth like a Zoomer on Netflix? Or having X-forwarding at all, for that matter? Fuck you, Wayland), and so fucking much more.
God that sounds even worse than Unix already was.
>ifconfig should have been updated, not replaced by the new hotness
it was updated, and that was actually a significant factor in many distros moving to iproute2, because that update changed the output format and broke a bunch of scripts. I also find it funny that you advocate for keeping the old stuff and updating it for ifconfig, while at the same time ranting about how we ought to instead be throwing everything out and starting over.
Also text is absolute the only sane choice for pipelines, because its the simplest thing you can do. Everything understands plain text, it's human-readable, and we have tons and tons of tools for dealing with it. Oh but I suppose you want those thrown out and rewritten, too.
The amazing thing is that not only better _is_ possible, but it already existed long before Unix was a thing.
Unix is not worth saving anyway.
Proposing "enhancements" is futile when the foundation is rotten.
>Also text is absolute the only sane choice for pipelines, because its the simplest thing you can do.
No it's not. It forces you to serialize and re-parse all the shit you send in pipes for no reason, where on Lisp-based OSes such as Genera you could reuse that structured data directly.
Pipes are retarded and a workaround to justify the "Unix philosophy", which is superficially attractive because jt makes every single part LOOK simpler, but complicates the whole, which is what matters.
This man says the truth
There have been numerous attempts to change it (GoboLinux, macOS, Android) but none have caught on because they are equally as arbitrary. The only way forward is a flat database-driven filesystem.
fagOS didn't try to do anything to change it, it merely inherited it. Applel was a complete joke and realized their own programmers were literally brain dead and couldn't produce a stable OS to the point of bankruptcy.
mac and android have as basic design objective hiding all that stuff. they want the user kept in their retard-proof walled garden that sits on top of the actual OS.
>database-driven filesystem
isn't that what MS tried to do with WinFS in the 2000s? and after five years they couldn't get it to work properly and were absolutely drowning in complexity so they just shitcanned it? not a promising precedent for that idea there
Gobo linux is best.
Yeah, and that shit only works if all the programs structure their data the same way.
There is no data format that covers all cases among all languages, which is why using text and letting programs parse it into their own internal structure is the best way.
Unix FS hierarchy is the best gatekeeping ever existed.
Keeps retards like you out for good.
Sometimes I wonder if Eunuchs weenies even read the posts they reply to in the first place.
It's not "gatekeeping" when people who are both smarter and more influential than you point out the objective truth of the FSH originating for mere technical limits which are a non-issue today.
There is literally zero reason for the split among /bin/ and /usr/bin/ on most modern machines, for instance.
Yet it persists.
Ironic that the most ardent proponents of the "Unix philosophy" defend unnecessary complexity.
>The hierarchy, like most other things in Unix, was hardly "designed" at all. It merely originated out of historical technical reason.
>Implying that's bad
Oh, but there is. And we had it decades ago.
en.wikipedia.org
>Files and other persistent objects are recorded in a repository called the Catalogue. Unlike other operating systems, the file naming hierarchy is independent of the location of a file on a particular tape or disk volume. In days where there was more need for offline storage, this made it easy to keep track of files regardless of their location, and to move files between locations without renaming them. As well as files, the Catalogue keeps track of users and user groups, volumes, devices, network connections, and many other resources. Metadata for files can be held in an object called a File Description. The Catalogue was probably the first example of what would later be called an entity-relationship database.
A solution to this problem was found in the 1970s, but instead of fixing anything, the AT&T employees preferred to bureaucratize and standardize their mistake without any solution. If that /usr disk became full and they still needed more space, they would have had to pick another random directory. Maybe all this bullshit is why UNIX weenies think UNIX was the first time anyone attempted a hierarchical file system, because nobody else screwed it up this bad.
It _is_ bad.
Ask 10 people what the role of /usr/bin/ or /opt/ is, and you'll get 10 different answers. This prevents streamlining and organization, despite what Rob Pike has been telling you for years. There is no logic behind it.
The most amazing example is /etc/ being retroactively named "Editable Text Configuration" when it originally stood for "etcetera", I.E. all the shit we don't know where to put.
What really drives me mad is not so much about it being bad. It's about so mang people unconditionally defending it.
Nearly every core utility and functionality is written in C in the original Unix and current Unix-like systems, so language portability shouldn't be an issue in principle.
The problem is that C is an exceptionally bad language for general scripting and automation, so the system needs ANOTHER language, the shell, for this task.
But this now introduces the problem of passing data from different places to others (which are now forced into being separate processes, in order to have a separate language call them), and the only reliable way, at this point, is in fact serializing and parsing over and over, every time there is some communication involved.
But all this unnecessary work could have been avoided by using a better language for the OS in the first place.
Itt
Cute!
i love windows nt kernel because seeing the similar clusterfuck they made with win9x they did it right the second time.
NT no longer has lettered drives, everything is an object (some of it named (like files, drives, etc..), aka accessible from object system by name) which they kinda borrowed from unix, has a robust permission system on all objects, strict conventions.
I just wish they dropped the win32 subsystem autism that is forced onto usermode.
But i guess its easier for the users to remember "D:" is my pendrive than knowing that \Global\D: is just a symlink to \Disk\HarddiskPartition5
I prefer the systemd file hierarchy (man file-hierarchy)
Plan 9 and NixOS don't have this problem.
>everything is an object
>which they kinda borrowed from unix
Lel no, it's borrowed from real operating systems like VMS.
>I just wish they dropped the win32 subsystem
And break nearly every application in existence.
Seriously, why do Jow Forumsentoomen post things like this? Talk OS design without the slightest conception of backcompat?
(Me)
>autism that is forced onto usermode.
Should add that, even though you'll deliberately fail to argue to the point anyway.
>yet it persists
because it's not exactly a problem, is it?
backwards compatibility is aids desu.
they should've done this when switching to 64bit. support win32 on 32bit, only support ntapi on 64.
>backwards compatibility is aids desu
maybe it doesnt sound bad to you, but i'm the one who has to suffer and write hundreds of lines because some dumb nigger at IBM forgot to connect a memory line 40 years ago
Plan 9 is worse than snake oil. Plan 9 was created to make distributed computing look like snake oil. Like UNIX compared to Multics, it arrived years after working distributed computing protocols and is still worse. AT&T's "research" is inferior to what was already on the market and will always be, no matter how much time and money is put into it. It's "research" that results in an inferior OS, increases costs on every single level, decreases productivity, replaces simple solutions with complex problems (sometimes so complex that nobody believes they can be solved), makes things worse for users, and prevents real researchers from doing what works because of "compatibility" with horrible interfaces. UNIX and Plan 9 are anti-research that shit on 60 years of computer science. That's why UNIX technology like Plan 9, BSD, and Linux in 2019 is worse than what Multics did in the 60s. Plan 9 uses "tar" to copy directory hierarchies not for any "philosophical" reasons, but because the version of UNIX it was based on didn't have "cp -r" and they were not capable of adding it. It doesn't have dynamic linking because they weren't smart enough to copy Multics or another mainframe system that does it properly. It still has the same bc calculator and all that other UNIX bullshit. What sucks is when Plan 9 weenies point to something that UNIX does wrong and blame the entire thing itself instead of the broken UNIX implementation.
Lol, this copypasta is so wrong.
9P (Plan 9's network protocol) is literally the best protocol invented.
It's used by NixOS, Windows Subsystem for Linux, and QEMU.
Professional grade software sees it as the best solution for distributed filesystems. Fuck off with your disinfo copypasta.
Can you elaborate on why without appealing to popularity or using meaningless marketing buzzwords like "professional grade"?
This. This is also the reason why everybody should keep using Windows
Just look how Haiku does it.
HISTORY
A hier manual page appeared in Version7 AT&T UNIX.
git init ~ && git int /etc
Don't expect a response. When asked such questions, Bell Labs slaves fall into wintoddler tier arguments.
sauce on image?