I spent 18 years in the Linux console and I don’t regret it

It all started very prosaically — my grandparents bought me my first computer and invited a friend, who was a programmer, to come over. He showed to me how to use my first PC — and told me about another operating system called "Linux" that was popular with programmers and hackers. This information was stored somewhere in the back of my mind. Back then, I didn't have internet access, except for a 56 kB/s modem at school, to which I could use every 1-2 weeks for a few hours. So I couldn't download the Linux distribution to record it on a CD, and I had no idea I could even download it from the Internet. I got the Windows distribution from that programmer, and occasionally I'd rent other discs. Usually, there were no Linux distributions there — only repacks of GTA, Half-Life and other golden classics.

Back then, I was using Windows XP and spending a lot of time playing Counter-Strike with bots and playing through various famous single-player games. I already knew how to program in Pascal and I had the Turbo Pascal IDE and compiler, which I copied to a floppy disk from the school computer. Of course, I ran it in full-screen mode because it looked cool! Like in movies about hackers, where people did some kind of magic by typing in incomprehensible text.

All this idyll would have continued, but a couple of years later I saw on the shelf of a bookstore the book "Slackware/MOPSLinux for the user (with disk)". Of course, I quickly and decisively persuaded my grandfather to buy it. Since the computer was in my sole use, and I already knew how to reinstall Windows and make backups (mods for CS 1.6 and my saves for GTA Vice City) — it was time for some experiments! I had two CDs with Linux and Windows distributions, seven hundred and fifty man pages, 150 sheets of a self-instruction book with console commands and a lot of console utilities in /bin and /usr/bin, one 40 Gb hard drive, Xorg -configure which creates a non-working /etc/X11/xorg.conf, and a text 80x25 console. Not that all is needed to become a cool hacker, but once you start delving into the system internals, then follow your hobby to the end. The only thing that bothered me was fdisk. There is nothing more helpless, irresponsible and immoral in the world than a schoolchild using fdisk to partition a hard drive. And I knew that pretty soon we would dive into it.

At first I just read the book and learn the console commands. After Windows, it was awesome — you type some text, and the system answers you, just like in the movies. It doesn’t just draw an sandglass and it’s unclear what operating system is doing. After some time, I learned how to exit vim, manually configured the X server using man-pages and was able to install the Pascal compiler, with which I continued to learn how to programming using the ITMO University textbook. Then, I figured out how to compile programs from sources, which I received on disks from the Xakep magazine (Russian transliteration of the word "Hacker").

Of course, Windows XP still stayed in dual boot. The vast majority of the software I used was for Windows. Support for DOC and XLS files in Linux was bad in those days, and without GTA games I felt boring. And given the lack of access to the Internet, I couldn’t figure out what else to do with Linux — KDE 2 3 was installed, I tried all the programs from the MOPSLinux disk, read the chapters about DNS and the HTTP servers from the book to the holes, but didn’t see the point of using them on localhost without the Internet.

Looking back, I can say that the knowledge and skills I gained became the basis that I still use today. It turns out that it is very useful to be alone with Linux, when you only have access to a book, man pages and source codes for the programs you want to install — there is no Internet or friends you can ask. This allows you to learn a lot about the system and develop skills in solving a variety of problems.x

Now, if some old, mammoth-shit, but very necessary legacy software refuses to run, then that’s not my problem. I use ldd, make symlinks to the required versions of libraries and read the logs to start the program anyway. If the program crashes quietly, I'll use strace. If I hadn’t had these skills since then, I can’t imagine how much time I would have spent at work and in pet projects launching some not very "modern" and "right from the future" software.

Around 2007, I entered ITMO University and moved to St. Petersburg. This is how I got access to huge (compared to what I had seen before) bookstores — House of Books on Nevsky prospekt and DVK (House of Military Books), which was neither a house, nor sells military books. Since by that time I had already been indoctrinated by Xakep magazine — the following were swept off the shelves:

• "Linux in the original" from the publishing house BHV
• "Developing Applications in a Linux Environment" from Williams Publishing (the photo of the MOPSLinux disk above is in the background of pages from this book)
• K&R "The C Programming Language".

Since at that time I had the Internet via dial-up, by the cards¹, I rarely used it. Mainly I check de.ifmo.ru portal for students and download free as beer books from the Moshkov Library (Lib.Ru). When I tried to do something longer, the time on the card ran out, my Acorp Sprinter@56K hung up and I had to run to the post office again for a new card.

Therefore, again, as before, I sat with books, studied system calls, and wrote all sorts of simple programs to learn programming.

Unlimited Internet, which is familiar to us in the modern sense, appeared to me somewhere around 2009-2010 years. And then everything started to take off — I tortured the hard drive of my computer, installing various distributions there. Basically, I chose them according to the following principle: "Oh, what a beautiful desktop environment is in this distribution — let’s install it urgently!"

From that time I only have three screenshots left. Here I’m digging into the C code in GVim, covered with plugins (FluxBox window manager):

And here is some code in GNU Assembler:

At the same time, I mastered LaTeX, tired of problems with printing student reports, when a file made in Open Office was printed crookedly in a bookstore near the university. And so I could finally write the report text in vim and get a beautiful PDF output that looks and prints the same everywhere.

Well, I continued to master system programming. One of the first programs I wrote is still in the SVN repository on SourceForge. This is jabsh (https://sourceforge.net/p/jabsh/code/HEAD/tree/) — something like a jabber remote shell. I didn’t have the opportunity to get a static IP address at that time, but I wanted to do something on my computer remotely. At that time, I had a Siemens C75 with the Bombus Jabber client installed, in which I chatted in all sorts of Linux conferences on jabber.ru when I didn’t have a computer at hand. And then the idea came to me to write a daemon that would connect to the Jabber server, wait for console commands from me, execute them and send the execution result in a return message.

This thing even worked and I used it until I got a static IP address. I even had a user from India for whom jabsh for some reasons did not work, and we send e-mails to each other for some time about this.

Another one of the programs from those times is a summer project from my future scientific supervisor — termprogs, for managing a set of "terminals" through "workstations", with a central server where the whole things are connected.

Just at this point, I was finishing reading William Stevens’ book "UNIX: Network Application Development" and could put all my system programming knowledge into practice.

Somewhere in my 2nd or 3rd year at the university, I started taking classes² on system programming. At first we were taught to use the terminal and vim on thin clients from Sun Microsystems, with pot-bellied CRT monitors. During these classes, I do nothing for a whole semester — after all, I had already studied all this back in school. But then the fun began.

We spent half the semester studying regular expressions and the grep, sed and awk. But regular expressions passed me by and I used grep at the level: "well, if you pass it a string as a parameter, it will search for matches in the file".

And here are furious tasks, kilometers of regexps and all that jazz. By the end of the semester, regular expressions were falling off my teeth. Looking back, I can now say that this regexp course is another unshakable pillar that I still use constantly. I can’t imagine how much time and efforts my knowledge of regular expressions saved me.

---

I still don’t understand where idea "if you solving a problem and decide to use regular expressions, then now you have two problems" came from. My experience at work and at home shows that if you need to somehow cleverly parse a string using a regular expression, then you take sed or Java's Pattern and Matcher — and parse the string. Then you test the resulting code, send it to testers — and then it just works for years.

---

At the same time, I began to share my experience — write articles on welinux.ru, talking with other people on linuxforum.ru, and attend SPbLUG meetings. At one time I had a blog on WordPress, which I set up on some free VPS, which you could use as long as you did not go beyond the lower or upper limits on CPU and memory. That’s when I became addicted to writing all sorts of texts with amazing stories.

Around 2011, I made a fateful decision — to go into embedded programming. At that time, this area of ​​Computer Engineering seemed to me more interesting and romantic than "regular" programming. After all, there are no "simplifying" levels of abstraction here —you take and write code that works directly on the hardware! And then you debug the whole thing using blinking LEDs, debug printing via UART, an oscilloscope and other tools. And all the knowledge about bits, bytes, the internal structure of all kinds of EEPROM, SRAM and other things is used 24/7!

All relevant courses at the university included working with Windows — at that time the necessary development environments and compilers were mainly for this operating system. But, naturally, this did not stop me. For half of the software I used VirtualBox with Windows inside. For the second half, I was able to find the necessary native tools.

For training, we initially used special devices based on the MCS-51 family microcontroller. If the code for them could be written in anything — I used Vim/Emacs — then compiling and flashing the compiled binary into the device was more complicated. For compilation, I used sdcc, but for firmware, a special utility was needed — and m3p. It was written by one of the university professors in ancient times on the C language. Fortunately, this utility was written with cross-platform compatibility in mind, therefore, after a couple of minor edits in the source code, it calmly did its job under Linux.

In those days, having begun to get tired of all sorts of "modern" distributions, with their NetworkManagers, PsshPsshAudio PulseAudio, Avahi Daemon and other "innovations" were breaking my user experience, developed back in the days of Slackware — I came to Arch Linux. It was possible to quickly install a basic system without the programs described above, supplement it with only the software I needed, and calmly work and watch memes.

Then I already began to develop a certain set of software, which I constantly used. By an understandable coincidence ("Russian" Slackware as the first Linux distribution and my love to use the console like a "hacker") it was mainly console software:

• vim/emacs — for editing text and code.
• latex — for writing all sorts of complex documents, especially if they need to be printed or sent somewhere. Well, for drawing presentations, so as not to "get up twice".
• Some kind of tiling WM — anyway, after a month of using some beautiful KDE or GNOME, I came to the conclusion that by default all my windows were expanded to full screen and scattered across the desktops, depending on the window name. And since you can’t see the difference, then why waste disk space on a heavy DE, if I can get everything I need in some xmonad or i3wm? Although, all sorts of beauty in the form of shadows, animations and transparency, which DE gives me, please the eyes for the first couple of weeks, but then the "wow effect" is expectedly lost.
• Well, and all sorts of other console utilities with which I could work with or without consciousness: grep, sed, git, make, cron and so on.

Since then, I had a repository with dotfiles, in which I dragged my configuration files for the above-described programs from system to system.

Here are some desktop screenshots from those days. Here is xmonad on two monitors - urxvt on the left, Chromium on the right:

And here the editing of the student's report. On the left is the final document in apvlv, and on the right is the TeX source code in GVim:

After that, I tried many times to switch to a usual software with a GUI or to all kinds of Web applications, but it just didn't work for me anymore. There weren't many ways to customize it for yourself. Some configuration options were extremly limited. Or software wasn't as fast as I'd desired. Or it was just inconvenient — the startup focus wasn't where I'm used to seeing it, or the main window displays information I wasn't used to seeing, and so on.

The final straw was the "redesign" of GMail. It slowed down even more and required more RAM than before. At this point I switched to mutt. Luckily, this thing is not the subject to the mOsT mOdErN dEsIgN trends and its appearance doesn't change from year to year. It works pretty quickly, alas, it launches not so quickly, even with caching. This is because I've had all my emails in maildirs since 2009 (about 47 thousand emails).

But the main thing with mutt is that it won't change in one "fine" day at the request of the left heel of the someone in the design department at Google.

In general, I've come to see Linux as more of a practical tool than a something "religious". There's no longer the same level of passion around which people wage wars over which Linux distribution is best. It started to become just a convenient and familiar operating system for me, one I only needed a little of:

• Do not do anything critical, such as software updates, without my knowledge.
• Just execut ethe programs, which are familiar to me.
• Stick to FHS³ and other standard things that I have learned since the days of Slackware — so that if some problem suddenly arises, I can quickly and calmly figure it out, understanding what is going on in the system.
• Do not impose on me a way to store my files — everything should be sorted into directories in the way that is convenient for me, without any tags or stars with the file rating. Or predefined directories.

[init]
  defaultbranch = master

USE="-systemd unicode -pulseaudio X alsa"

Audacious on the left

It's evident that I'm not quite at the level of a "cool Linux hacker, committing patches to the kernel instead of breakfast, lunch and dinner" (yet). But all those years of tinkering with console utilities paid off. I ended up with a pretty stable and simple system that I can use event without consciousness. In this system no one app will change it's interface by itself according to "new fashion trends".

In this system all my settings are stored in Git, so nothing will change without my knowledge. I can do whatever I want with just a couple of lines in the desired file and a several basic commands combined via pipe, for example:

• The plain-text accounting utility didn't allow me to use the "cash envelope" system I was used to. I used dialog, awk and sqlite3 to create a budgeting system on top of hledger that does everything I need.

• I bought myself a Logitech Trackman Marble trackball, which has the "Forward" and "Back" buttons I don’t need. But there is no middle mouse button or scroll. And this is not a problem.

  I create a file /etc/X11/xorg.conf.d/50trackball.conf with the following lines:

  Section "InputClass"
      Identifier "Marble Mouse"
      MatchProduct "Logitech USB Trackball"
      Option "EmulateWheel" "true"
      Option "EmulateWheelButton" "9"
      Option "MiddleEmulation" "true"
      Option "ButtonMapping" "3 8 1 4 5 6 7 2 9"
      Option "XAxisMapping"  "6 7"
  EndSection
  

  The "Back" button now works like the middle mouse button, and if hold down the "Forward" button, I can scroll the text in all directions with the ball. As I'd wanted, the trackball is now left-handed.

• The new keyboard has Fn buttons for "My Computer", "Search", and "Browser", but no buttons for volume control? No problem! I use xmodmap to reassign the button codes in the file it generates:

  keycode 152 = XF86AudioLowerVolume NoSymbol XF86AudioLowerVolume
  keycode 163 = XF86AudioRaiseVolume NoSymbol XF86AudioRaiseVolume
  keycode 180 = XF86AudioMute NoSymbol XF86AudioMute
  

So, for me, Linux is now just a system that runs the programs I'm used to — which, like a wall made of bricks, form my familiar user environment. The bastions — represented by Gentoo and Devuan⁵ — are currently protecting me from the overwhelming ​​changes that aren't necessary for me and related problems. While the rest of the Linux world is changing the initialization systems, moving away from the X server and rewriting coreutils in Rust, I'm still using the same tools I've always used. I'm just easily read email and RSS feeds in mutt year after year.

When (if) these bastions fall, I’ll probably move to the monastery to FreeBSD. Luckily, I've already got some experience using it as a regular user. All my other software, like i3wm, emacs, Firefox, RawTherapee and so on, also works there. The only big changes in my configuration that will have to be made are to call gmake instead of make in some Makefiles, and to use more correct she-bang #!/usr/bin/env bash in scripts, instead of the usual #!/bin/bash. Unfortunately, I'll have to say goodbye to Docker, which isn't available on FreeBSD, and the ability to work with LUKS crypto-containers. But it’s better to lose them than all my familiar, configured with love environment and my long-term habits.

My entire history of mastering Linux can be described as "hard to learn, hard to master". But over time, I developed all kinds of different habits that let me write texts, use the internet, and so on — literally "at my fingertips". That's why I’m not here advocating you immediately switch to i3wm or Emacs for the sake of pRoDuCtIvItY. Without the habits I've mentioned, it won't help. First of all, you have to want to learn, for example, Emacs, and be prepared for the fact that you will have to configure it for some time, and not perceive this time as "well, I need to set up a text editor instead of just opening it and writing text" — and then something will work out. I think all these articles about switching to Vim to be more productive in programming are misleading. Firstly, you'll spend time on vimtutor instead of programming. Secondly, there's not a strong connection between typing speed and programming. I can type at a speed of only 60-70 characters per minute, but this doesn't affect my productivity as a programmer. After all, I'm typing code on the keyboard at most 20-25% of the time. About 10-15% of the time is spent communicating with colleagues and on Zoom calls to figure out what's going on with this task or bug. The other 60-70% is spent reflecting in front of a notebook with a pen in hand, thinking: "how can I make a change correctly and quickly, so I don't waste a lot of time, either now or in the future?" So, vim won’t help with productivity here — it doesn't think for me in front of a piece of paper.

Third, let's be real. Right now, for a lot of languages, a big, complex IDE is still a better choice than Vim or Emacs. Even if you have an LSP server⁶ for your editor. For instance, Emacs' LSP for Java still doesn't work very well — it crashes on simple things, it doesn't update the context of changes in files as quickly as IDEA does, and requires some finesse to make it work with Lombok.

As a general rule, you can get a lot done in the GUI and it's the best place for it. It's best to develop photographs in RawTherapee, edit images in GIMP, view the site in Firefox, and so on. But there are lots of other actions you can do right from the console. It's just matter of convenience. Some people find it easier to select files to copy with the mouse in Nautilus, while others prefer to use the cp ~/photos/{photo,video}_*.{jpeg,jpg,JPG,avi} /media/BACKUP. It's great that Linux (for now) offers a choice for both people who are used to a graphical interface and for those who prefer to communicate with the machine by text.