The Renaissance of the Shell?
effective-shell.com
effective-shell.com
Of course this is driven by commercial interests. Microsoft tries to close the gap to the rich Mac OS X development environment, which is rich because it is compatible to the Unix/Linux ecosystem.
As the author noted, many jobs are primarily terminal-based in a way like probably nobody would have believed 20 years ago. DevOps or data scientist, they both enjoy powerful CLI applications, REPL interfaces, scripting as if the "GUI movement" did not even take place.
VisualAssist was a very popular plugin for it (don't know if it's still relevant) which added functionality inside the GUI.
For instance: A typical 2D drawing program which would primarily be used through the UI to draw images could also register new image format plugins with the operating system, so that all applications that deal with image files suddenly can load and save those new formats. The same application might be started on the command line in a 'headless' mode to offer image manipulation tasks like ImageMagick, and finally the application can offer an AREXX scripting port for more complex automation tasks, and "orchestrating" several applications.
All of this was standardized through "best practices" guidelines by the Amiga team.
I miss the Amiga :)
PS: My pet theory for the reason why there's a "shell renaissance" is that UI usability (in the sense of making complex application features "usable") has dramatically regressed in the last 10..20 years because the target audience for most apps has changed from active users to passive consumers. People who don't fit the "average user" have no other choice than to move to the command line because UIs are no longer designed for them. This "average user" targeted by UI designers is no longer a creative person who uses the computer to solve tasks or create art, but a passive media consumer. The computer has been degraded from the "bicycle for the mind" to a dumb TV set basically.
Tyranny of the Minimum Viable User
https://old.reddit.com/r/dredmorbius/comments/69wk8y/the_tyr...
The command line tools are simply more versatile and configurable.
We can try to take git clone command as an example and try to design a GUI dialogue which would allow to use all the options it has.
I think it is very hard and you will end up with something looking scarier than space shuttle cockpit.
I believe it'll take roughly the same time to make sense of that dialog that it takes me to quickly man git-clone to remind myself about some obscure option I don't use everyday.
I don't hate on GUIs, I just think that it is very hard to design full-featured GUI for complicated tools that we use every day. I would like to see an attempt to do this for Git, the key is being full-featured.
Prior to trying it I was a hardcore git terminal user. I still drop to the terminal sometimes, but magit has won me over for the day-to-day normal workflows.
You have to use Emacs for Magit, as it's an Emacs Lisp package, but it really is worth trying in its own right.
(I actually used this feature recently when I wrote a Blitz Basic plugin for modern Emacs. It uses ARexx over the network to talk to Blitz, so you can write code in your modern operating system, and then compile and run on the Amiga running alongside in an emulator[0])
There were also a number of visual programming environments such as Amiga Vision and Scala which were marketed as "Multimedia Authoring Systems". They were used to create information kiosks, advanced interactive presentations and early CD-ROM hypermedia products such as multimedia encyclopedias, reference works and the like. I'm surprised there's nothing like them today, they seem ideal for the simpler end of the mobile ecosystem, or more cynically for those who want to turn their whole platform into unhackable (in both senses of the word) information kiosks.
Back in the 90s I looked at systems like this, or at Visual Basic, or NeXT Interface Builder (which later made its way to Mac OS X) and thought that the future would be some extension of these technologies, not the tottering stack of cli tools we use today.
I too miss the Amiga, and consider myself a proponent of desktop personal computing and "bicycles for the mind", which can be a depressing thing to be in these times.
If it is anything like Powershell, which is more "modern" than bash, then I don't think it will make much of an impact. Powershell seems like it was designed by a comittee which did not quite understand the point of what they were doing.
It seems that if at least someone on that committee used a Unix terminal maybe once in their lives they would realize that what they were building would deliver a really bad user experience.
They (Microsoft) say that their inspiration was Korn Shell and Languages like TCL. Microsoft when they still offered Windows services for Unix, they did provide a Korn shell. If you look at some PowerShell examples and some Korn Shell example there are quite a few similarities between the two.
So they obviously were quite aware of the options in the *nix world. I really wish people wouldn't make these sorts of claims when spending like less than a few minutes looking up the development history would dispel these ideas immediately.
Personally I really like PowerShell and I find it fun to program in whereas with bash I seem to have to always relearn it whenever I attempt to do something non-trivial. However I come from a background of programming in OOP languages like Java and C# so many it just suits my mental model better than bash.
Using a powershell cli is very verbose (although pretty consistent), but using it without some sort of GUI helper if you're not familiar with it is pretty daunting.
Also there are lots of extensions that you are kinda just supposed to know about e.g. dba-tools extensions.
And then there are other advantages too. Editing code in Vi/Vim is akin to touch typing if you are really good. Programming becomes so much easier and so much fun.
And last but not the least, there are still loads of things to learn about Vi.
I did switch from vim to Emacs + Evil though (Evil adds a decent text editor to Emacs), with my own set of spacemacs-like keybindings.
(1) Systems being written in more languages means the shell becomes more important
I wrote about that here: https://news.ycombinator.com/item?id=24083764
The idea is that I write programs in Python, JavaScript, R, and C++ regularly, and about 10 different DSLs (SQL, HTML, etc.) And I work on systems written by others, consisting of even more languages.
in regards to http://www.oilshell.org/
(2) Convergence around Unix and Linux for the server side.
Unix won. In 2017 or so, Windows Subsystem for Linux was marketed as "Bash on Windows" !!! That is, being able to run a 30 year old shell and other programs was a new feature of Windows.
(3) DevOps
I forget where I read this, but someone quipped that "old school sys admins didn't disappear". (That is, the experts at Unix shell.) "What happened is that they went to work for AWS and Google and then sold their skills back to you at a higher and recurring price"
I find that to be pretty spot on... The cloud companies are making a lot of stuff point and click, and cut and paste YAML, so you don't have to use shell, but I think programmers are better off learning and using shell.
Reaosns: For autonomy, to avoid being locked in, to exercise the ability to create simple, sharp, one-offs ... not drag in a 200 MB cloud SDK to solve a simple problem.
I have been using Nix for most system-level stuff the last two years (development environments, building container images, declaratively managing systems). It is far more powerful than YAML configuration (it's a turing-complete functional language after all), but the functional aspect brings many benefits (no side-effects, etc.).
Of course, nix(pkgs) stdenv makes liberal use of Bourne shell in its build phases, so a bit of shell chops is very useful there as well.
A picture is worth 1000 words.
Why use 1000 words to do something when one or two will do?
One evening, Master Foo and Nubi attended a gathering of programmers who had met to learn from each other. One of the programmers asked Nubi to what school he and his master belonged. Upon being told they were followers of the Great Way of Unix, the programmer grew scornful.
“The command-line tools of Unix are crude and backward,” he scoffed. “Modern, properly designed operating systems do everything through a graphical user interface.”
Master Foo said nothing, but pointed at the moon. A nearby dog began to bark at the master's hand.
“I don't understand you!” said the programmer.
Master Foo remained silent, and pointed at an image of the Buddha. Then he pointed at a window.
“What are you trying to tell me?” asked the programmer.
Master Foo pointed at the programmer's head. Then he pointed at a rock.
“Why can't you make yourself clear?” demanded the programmer.
Master Foo frowned thoughtfully, tapped the programmer twice on the nose, and dropped him in a nearby trashcan.
As the programmer was attempting to extricate himself from the garbage, the dog wandered over and piddled on him.
At that moment, the programmer achieved enlightenment.
Source: http://www.catb.org/~esr/writings/unix-koans/gui-programmer....
ls -al
rm -r. /*
git checkout -b thisIsFine
I read the parable as being about the infinite number of ways icons and GUIs can be interpreted, thus making their inherent discoverability and clarity low.
I don't see how listing CLI commands with terrible UX invalidates that point.
Icons don't have an intrinsic meaning the way words do, so some percentage of people will find brand-new interpretations for icons, to a larger extent than will happen with words.
That was my interpretation, at least.
This is completely backwards. Icons can have intrinsic meanings - they can literally depict an object or an action from the real world. Words are pure abstracts humans made up, and can not exist outside of the context.
Same goes for google-fu. What is happening that the arcane science of proper search semantics is not being taught kids in school? It seems to me the devolutionary effect of "ease of use computing" has wrought its anti-pattern woes over a few generations ..
Anyway, my kids learned bash before they learned how to find the Settings panel, and there is something to be said for the teenager who knows how to wrangle his OS because his Dad taught him the basics of shell-based package management, something still not quite done well in certain environments...
I sometimes wonder if how people talked about 'this whole writing and reading thing' way back when. Perhaps there's a future where the average person will be what we now consider a 'power user'?
Much as there's the criticize about current-day computer use, as someone who was considered a 'nerd' for chatting online in my teens, it's still amazing to see extremely non-typical 'nerds' sit at their laptops or phones chatting with others. Or to have conversation about gaming with people who fit the 'jock' stereotype.
I'd love to hear from historians how the 'common people' thought about writing back in the day when most people were illiterate.
I guess we (although I am not the person you replied to) have a fundamental disagreement here. I think absolutely, everyone is able to code, a little. Not to be a career programmer, mind you, and I have no expectation that they should be able to produce readable or maintainable code. I have a weakly-held opinion that people shouldn't have to code if they choose not to. However, I think a little coding knowledge is essential for digital autonomy, which is more and more synonymous with regular autonomy, and that everyone ought to have that.
The world seems to be moving much more towards APIs and declarative configuration formats, with actual scripts acting as a sort of last resort only. The author even mentions Kubernetes, which is very much designed to replace scripting based control of your systems with an API. Even kubectl is just an utility build over the API, not a core part of the system.
And even outside the cloud-ish world, you have systemd taking over the Linux world one distro at a time, and it's focus is firmly towards replacing shell scripts with configuration files.
Shell should be considered basic and most important interaction with computer with anything else deemed as optional.
The big problem is that almost all of the shell's are stuck in the previous century, except PowerShell.
Another exception is Nushell, which seems to be partly inspired by PowerShell, and has some very appealing ideas. However, given the time Fish needed to steal even a small market share from Bash and Zsh, it will likely take a long time before something even more radical goes mainstream. Unless of course a major player were to back this new shell, like when Microsoft pushed PowerShell and Apple pushed Zsh.
I was talking about production ready systems. Nushell is far from it ATM, and will certainly be so in the next half decade.
Zsh for many, many years was basically Bash with more features but not endorsed by GNU.
Bash was released in June 1989.
Zsh was released in December 1990.
Yet Zsh has replaced Bash only on MacOS, only about 1 year ago and only because of Bash licensing issues (new Bash versions are under GPL3 instead of 2 and Apple hates GPL3 because it's against their locked down devices).
29 years to replace the default shell on just 1 platform... despite almost bug-for-bug compatibility, POSIX support, extra features.
I wish new shells good luck!
Is it? Or is it that us compsci and computer engineering folks aren't used to it and don't know much about it?
A Japanese instructional YouTuber I watch did a tour of his setup, and in it he shows off a lot of automation he uses to make his work faster, more consistent, and less error prone. The relevant part starts around here and lasts throughout the rest of the video (its in english): https://youtu.be/TGgbSmCoZFA?t=390
I think part of it is there is already a buttload of CLI driven tools for us, so there isn't a need for us to rise above the skill floor for GUI automation. But to just assume it is hard or impossible because we don't know is just the same as people new to CLIs saying they are hard or impossible.
The GUI is heterogeneous world. Linux doesn't have tools AFAIK work mentioning because it didn't need them due to CLI. Tools like xdotool are very limited.
Windows had much better tools, such as AutoIt, AHK etc. This was because they all were created prior to PowerShell when there was no other option that encompasses all apps. And PowerShell doesn't deal with GUI too, although you could force it.
However, big part of those scripts depend on current theme, window sizing, even positions. To create good script is not easy.
Until a better way comes out to tell the computer what to do, the command line will prevail.
(The last few years have not been too kind to this metaphor because of computer-based analytics, but still)
People care about significant whitespace because it's inferring heavy semantic meaning from something that's by default invisible, and even when shown visibly, may not actually be particularly obvious.
When I code in C and move some code around I can just tell my editor "reindent this code" and it looks fine immediately. With Python I always have to double-check to make sure everything is in the right place.
Similarly I almost never have to reindent anything manually in C. The editor always knows where I should be based on the number of open braces. In python I often have to readjust. A common situation where that's annoying is if I want to add code after and outside a block, in C I just tell vim to open a line after the closing bracket and I'm immediately at the right location, in Python doing this will have the cursor at the wrong level of indentation.
It's not the end of the world and it's bikeshed territory but for me significant whitespace is in the same category as automatic semicolon insertion in JavaScript, I sorta get why somebody thought it was a good idea at some point but it's just more trouble than it's worth in practice.
Python wanted to be the anti-perl and it went too far in some places. Pseudo-code is only "pseudo" for a reason.
Also I feel like changing the indent of the cursor (manually) when exiting a block is more or less the same work as typing }, and I got used to it to the point it becomes automatic.
But I guess at the end of the day it's a matter of priority. Is that more important than having special syntax for dicts? I really like the fact that we have the tuple/list/dict distinction, and would be great if we had a different paren type for sets (which would have eliminated the {} vs set() issue). On the other hand, going with begin..end or if..fi blocks would be even more antagonizing than using whitespace.
I definitely agree, I think that POSIX-shell scripts can often be completely unreadable and difficult to maintain. But I think the concept of scripting in itself is fine, even though the implementations we might use today are slightly outdated.
At the moment there's a huge number of interactive shells like nushell and xonsh being built, but they don't really focus on scripting; I'd really love to see more competitors attempting to take on the mess of bash scripts.
The same could be said for any language, if you're not familiar with it.
For instance: you can't even slice "$@" which is the only array like construct specified by POSIX (excluding unsafe splitting), so you end up shifting and re-enqueuing with set -- "$@" "$1", which is unreadable.
Want to read a single character? Impossible with the POSIX read, but you can workaround with dd
etc.
All those workarounds to POSIX limitations are fascinating, but it forces a lot of arcane constructs, that's for sure.
And of course this is usually the point where you ask yourself if you've chosen the correct language for the task, but that's another debate.
How is that unreadable? Set all the positional arguments as is, then set the first one again, into new positional arguments. I'm not sure why you want the first arg twice, but thats your call.
Either way, it's hardly unreadable.
> And of course this is usually the point where you ask yourself if you've chosen the correct language for the task, but that's another debate.
It isn't another debate though. It's all the same debate: do you know the language you're reading and/or writing? Do you know what it can do, what it can be pushed to do, and what it really shouldn't do?
Those are all related to the same basic point: if you think Shell is unreadable, I'd suggest it's because you don't know shell.
Remember readable means that you can read and understand it. Not that it's written out in words that someone with on page 1 of "how to program for dummies" can understand.
Yes you are right, my usage of "readable" was probably wrong there.
Maybe the point I was trying to make was less about the individual constructs but more about how the lack of "common" features makes the whole program less "understandable", or maybe less easy to get familiar with, simply due to the amount of code needed to achieve a specific task.
The same way assembly is considered less "readable" than C. Not because assembly is less readable on a line by line basis, it's even simpler, but because of the number of lines and operations needed to achieve a simple task.
Basically it's easier to understand 10 lines than a 1000.
Check out http://www.oilshell.org/ -- that is exactly its purpose. It runs your bash scripts, and then you can upgrade gradually to a better language.
It's also an interactive shell, but I concentrate on the language first.
Syntax like var=value and [ x -eq y ] was super weird coming from C and Python.
My Oil project fixes many of these problems:
http://www.oilshell.org/blog/2020/01/simplest-explanation.ht...
(as well as fixing semantic problems; it's not just syntax)
Looking at modern desktop Linux, I wish it had retained more of this. Unfortunately, d-bus, GObject and Wayland aren't really file-based.
Recently there was a thread about Factorio (the game), which is all about automating. And once you have standard-ish inputs & outputs, you can achieve a lot of automations, in a different way.
It also highlights that the shell has limitations: it's hard to give multiple inputs, and handle outputs and create a complex pipeline. (I guess that's why we have things like Airflow)
Or perhaps even included the full name: "The Renaissance of Royal Dutch Shell".