Payment by bank transfer · Moroccoتعلَّم · مارِس · أتقِن

Free guide

From UNIX to Linux: The History That Explains Everything

The story that explains why Linux works the way it does, in 7 chapters. Read online or keep as a PDF.

Follow the course

Chapter 1

From Multics to UNIX (1964-1973)

Objective

Understand how UNIX, the ancestor of Linux, was born, and why its creators' ideas, more than fifty years old, still explain the way you will work on a Linux server today.

Key points

  • Multics (started in 1964) was a very ambitious operating system project; AT&T's Bell Labs withdrew from it in 1969.
  • That same year, Ken Thompson wrote a small system on a PDP-7 computer: this was the beginning of UNIX.
  • Dennis Ritchie created the C language; in 1973, the core of UNIX was rewritten in C, which made it portable from one machine to another.
  • Pipes (|) arrived in 1973: you combine small programs to do big things.
  • The UNIX philosophy — "a program does one thing and does it well", "everything is a file", "text as the common language" — is still at the heart of Linux.

1. The context: computers the size of a room

In the early 1960s, a computer filled an entire room, cost millions and served only one person at a time. You prepared your program on punched cards, handed them in at the counter, and came back for the result… the next day. One missing comma, and the day was lost.

The idea that excited researchers at the time was called time-sharing: the computer switches very quickly from one user to the next, so quickly that everyone feels they have the machine to themselves, at their own terminal.

This is exactly what a Linux server still does: dozens of people and hundreds of programs use it "at the same time". When you type who or ps in this course, you will see the direct legacy of this idea.

2. Multics: the big project that disappointed (1964-1969)

In 1964, three partners launched Multics (Multiplexed Information and Computing Service):

Partner Role
MIT (Project MAC) University research
General Electric The hardware (the GE-645 computer)
Bell Labs The research laboratory of AT&T, the American telephone company

Multics wanted to do everything: serve hundreds of users, protect every file, run without ever stopping, like the power grid. Several of its ideas were excellent: a directory tree, access rights on files, a command interpreter that is a program like any other.

But the project fell behind schedule and became heavy and expensive. In 1969, Bell Labs decided to withdraw. Multics continued elsewhere (the last Multics system in service was shut down in 2000), but the Bell Labs researchers found themselves without a project… and with their heads full of good ideas.

3. 1969: Ken Thompson and the PDP-7

Ken Thompson wanted to keep working on a system of his own. He picked up a PDP-7 from the company DEC, a minicomputer that was already outdated at the time and was sitting unused in a corner of the laboratory.

It is often said that Thompson also wanted to run a game he had written on it, Space Travel, a simulation of the solar system. For that, he needed a file system, a way to launch programs, an editor and a command interpreter. In the summer of 1969, he wrote all of this in a few weeks, in assembly language.

He was soon joined by Dennis Ritchie, then by other colleagues: Doug McIlroy, Joe Ossanna, Brian Kernighan…

Where does the name come from? A colleague (usually credited as Brian Kernighan) suggested "Unics", a nod to Multics: where Multics wanted to do everything in "multi", this small system did things simply. The spelling became UNIX.

A detail that will be useful to you: UNIX and Linux systems count time in seconds since 1 January 1970 at 00:00 (UTC), a date chosen at that time. It is called the "UNIX epoch". Try date +%s: the large number displayed is the number of seconds elapsed since that day.

4. From the PDP-7 to the PDP-11: UNIX finds its place (1970-1972)

To get a more powerful computer, the team made a promise to management: to write a text-processing system for AT&T's patent department. In 1970-1971, UNIX moved to a PDP-11, a much more common computer. The patent department became the first "customer" of UNIX.

In November 1971, the first UNIX manual was published (the First Edition). The following editions came out every few months. The format of this manual — one page per command, with NAME, SYNOPSIS, DESCRIPTION — is the format of the man pages you will read on Linux.

5. The C language: the decision that changed everything (1972-1973)

At the time, an operating system was written in assembly language, the language specific to each processor model. As a result, a system written for the PDP-7 worked only on the PDP-7. Changing computers meant rewriting everything.

The lineage of the C language looks like a family tree:

BCPL (Martin Richards, Cambridge, 1966-1967)
  └── B (Ken Thompson, 1969-1970) — simplified version of BCPL
        └── C (Dennis Ritchie, 1971-1973) — B enriched with data types

Ritchie created C in the early 1970s. It is close enough to the machine to write a system, but general enough to be compiled on other processors.

In 1973, the team rewrote the UNIX kernel in C (the Fourth Edition, November 1973). It was a first at this scale: only a small part remained in assembly language. From then on, porting UNIX to a new computer mostly meant writing a C compiler for that machine.

This decision explains why Linux runs today just as well on an Intel server, an ARM phone or a supercomputer: its kernel, too, is written mainly in C.

In 1978, Kernighan and Ritchie published The C Programming Language, the reference book that programmers still call "K&R".

6. Pipes: Doug McIlroy's idea (1973)

Doug McIlroy, the head of the team, had been asking for years for a way to connect programs to one another, like garden hoses. Thompson implemented it in one night, in 1973 (Third Edition): this is the pipe, written |.

Example: you want to know how many users on your server use the bash shell.

# Without a pipe, you would need a special program for this specific question.
# With a pipe, you combine three small tools:
cat /etc/passwd | grep bash | wc -l
#   cat  : reads the accounts file
#   grep : keeps the lines that contain "bash"
#   wc -l: counts the remaining lines

None of these three programs was written to answer this question. Together, they answer it. That is the whole strength of UNIX.

7. The UNIX philosophy, in four rules

These principles were summarised later by McIlroy and others; they remain the best compass for understanding Linux.

Principle What it means Example on Linux
A program does one thing, and does it well Small, specialised tools rather than one giant piece of software ls lists, sort sorts, grep searches
Programs work together The output of one becomes the input of another ps aux | grep httpd
Text is the common language Configuration and exchanges in readable text /etc/passwd, /etc/hosts, log files
Everything is a file Disks, terminals, kernel information: you read and write them like files /dev/sda, /proc/version, /proc/cpuinfo 
cat /proc/cpuinfo     # the processor, presented as a simple text file
cat /proc/version     # the kernel version, same idea
ls -l /dev/null       # a special "file" that swallows everything you send to it

For the administrator, the consequence is very concrete: almost everything is configured by editing text files, and almost everything is diagnosed by reading text files. No graphical tools needed.

8. 1973-1974: UNIX leaves the laboratory

In October 1973, Ritchie and Thompson presented UNIX at an academic conference on operating systems (SOSP). The paper was published in July 1974 in the journal Communications of the ACM. Universities discovered this elegant system and asked to try it.

However, because of an antitrust agreement signed in 1956 with the US government, AT&T was not allowed to do business in computing. So it distributed UNIX to universities for a nominal fee, with the source code. Thousands of students would study it, modify it and improve it. This is the starting point of the next lesson: the large UNIX family.

In 1983, Thompson and Ritchie received the Turing Award, the equivalent of the Nobel Prize in computer science, for UNIX. Dennis Ritchie died in October 2011; Ken Thompson later took part, at Google, in creating the Go language.

9. The lesson timeline

1964  Launch of Multics (MIT, General Electric, Bell Labs)
1969  Bell Labs leave Multics — Thompson writes UNIX on a PDP-7
1970  Move to the PDP-11 — the name "UNIX" takes hold
1971  First UNIX manual (First Edition, November)
1972  Birth of the C language (Dennis Ritchie)
1973  Pipes "|" — the kernel rewritten in C (Fourth Edition)
1974  Paper in Communications of the ACM: UNIX becomes known
1983  Turing Award for Thompson and Ritchie

Common misconceptions

  • "UNIX was designed by a large company to be sold." No: it was a project of two researchers, almost clandestine at first, on a salvaged machine. AT&T was not even allowed to sell it.
  • "Linux is a version of UNIX." Not exactly. Linux resembles UNIX and follows its rules, but it contains not a single line of code from UNIX: it was rewritten from scratch (lesson 0.4).
  • "Everything is a file" means everything is stored on the disk. No: /proc/version does not exist on the disk. The kernel generates it on the fly when you read it. It is the way you access it that is the way you access a file.
  • "C is an old, abandoned language." The Linux kernel, most of the basic tools, and databases such as PostgreSQL are written in C. It remains one of the most widely used languages in the world.

Lab

Your first machine is waiting for you in the terminal, running Rocky Linux 9. You will make it talk, using the habits inherited from UNIX:

uname -s                  # the name of the system
cat /proc/version         # "everything is a file": the kernel describes itself in a file
cat /etc/os-release       # the identity card of the distribution
cat /etc/os-release | wc -l   # a pipe: read, then count

Complete the tasks: save the name of the system, copy the kernel description, count lines with a pipe, and find the year UNIX was born. Then click Check: the site checks your work automatically.

Quiz

  1. What were the three organisations behind Multics?
  2. On which computer did Ken Thompson write the first version of UNIX, and in what year?
  3. Why was rewriting the UNIX kernel in C (1973) decisive?
  4. What does the command cat /etc/passwd | grep bash | wc -l do, and which UNIX principle does it illustrate?
  5. Why did AT&T distribute UNIX to universities instead of selling it?
Answers
  1. MIT, General Electric and Bell Labs (AT&T).
  2. On a DEC PDP-7, in 1969.
  3. Written in C, UNIX became portable: to run it on another machine, you essentially needed only a C compiler, instead of rewriting everything in assembly language.
  4. It counts the system accounts whose line contains "bash". It illustrates combining small specialised programs connected by pipes.
  5. A 1956 antitrust agreement prohibited AT&T from doing business in computing; so it distributed UNIX for a nominal fee, with its source code.

Reference sources: papers by D. M. Ritchie and K. Thompson cited in the header, Debian Reference (free licence). Original MasterLinux text.

Chapter 2

The UNIX family — BSD, System V and the POSIX standard

Objective

Understand how UNIX split into several rival branches, why a common standard (POSIX) became necessary, and why most of what you learn on Linux also works on macOS or FreeBSD.

Key points

  • At the University of Berkeley, UNIX became BSD: this is where TCP/IP networking, the vi editor and the csh shell came from.
  • After the break-up of AT&T (1984), the company sold its official version: System V.
  • Each manufacturer created its own paid UNIX (Solaris, AIX, HP-UX…), incompatible with the others: these were the "UNIX wars".
  • The POSIX standard (1988) defines what a "UNIX-like" system must provide; Linux follows it closely.
  • A lawsuit (1992-1994) blocked BSD at the very moment Linux was born, which pushed many developers towards Linux.

1. Berkeley receives UNIX (1974-1977)

As you saw in lesson 0.1, AT&T distributed UNIX to universities with its source code. The University of California, Berkeley installed it in 1974. Ken Thompson even spent a sabbatical year there (1975-1976) as a visiting professor.

A student, Bill Joy, started improving the system. In 1978, he released a first collection of programs: the Berkeley Software Distribution, or BSD. At first, it was a simple add-on to AT&T's UNIX; little by little, it became a complete system.

What BSD brought, and what you will use on Linux:

Berkeley contribution What you will do with it on Linux
The vi editor (Bill Joy, 1976) vi and its successor vim are present on almost every server
The csh shell Less used today, but command history and job control (jobs, fg, bg) date from this period
The TCP/IP network stack and sockets (4.2BSD, 1983) All modern networking: ssh, web servers, the entire Internet
The Fast File System (FFS) The idea of blocks and inodes taken up by ext2, ext3, ext4

The arrival of TCP/IP was a turning point: the US military agency DARPA, which funded the future Internet, chose BSD as its reference system in the early 1980s. UNIX and the Internet grew up together.

2. AT&T enters the market: System V (1983)

In 1982, an agreement with the US justice system provided for the break-up of AT&T, effective on 1 January 1984. In exchange, the company finally gained the right to sell computing products.

AT&T then marketed its official version of UNIX: System III (1982), then System V (1983), with its successive "releases" (Release 2, 3, 4). SVR4 (System V Release 4, late 1980s), developed with Sun, brought together a large part of the ideas of BSD and System V.

Several Linux habits come from System V:

ls /etc/init.d/       # the "System V style" startup scripts (RHEL 6, old Debian)
ps -ef                # the System V syntax of ps (with a dash)
ps aux                # the BSD syntax of ps (without a dash) — Linux accepts both!

The fact that ps accepts two styles of options is a living reminder of this double origin.

3. The commercial UNIX systems (1980s-1990s)

Every major manufacturer took a UNIX licence and created its own version, adapted to its machines (and sold with them):

System Vendor Base Still used?
SunOS then Solaris Sun Microsystems (acquired by Oracle in 2010) BSD, then System V R4 (Solaris, 1992) In a few large companies
AIX IBM (from 1986) System V Yes, on IBM Power servers (banks, insurance)
HP-UX Hewlett-Packard System V End of life
IRIX Silicon Graphics System V Discontinued (2000s)
Xenix Microsoft, then SCO System V Discontinued
Ultrix, Tru64 DEC BSD, then OSF/1 Discontinued

The case of Sun is interesting: the company was co-founded in 1982 by… Bill Joy, the author of BSD. Sun workstations were the reference machines of universities and of the Internet in the 1980s and 1990s.

4. The "UNIX wars" (1988-1993)

Each of these UNIX systems added its own commands, options and tools. A program written for AIX had to be modified to run on HP-UX. An administrator trained on Solaris was lost on IRIX.

At the end of the 1980s, two camps fought each other:

          AT&T + Sun                       IBM, DEC, HP and others
   (System V Release 4, 1988-1989)       (Open Software Foundation, 1988)
       "UNIX International"                       "OSF/1"
                   \                               /
                    \___ incompatibilities, lawsuits, ___/
                          lost customers

While the manufacturers were fighting, a competitor was quietly advancing on PCs: Microsoft Windows NT (1993). And, without anyone suspecting it yet, a Finnish student started writing a kernel on his PC in 1991…

5. POSIX: the common rule (1988)

To get out of the chaos, the standards body IEEE published the POSIX standard (IEEE 1003.1) in 1988. The name, suggested by Richard Stallman (whom you will meet again in lesson 0.3), roughly means Portable Operating System Interface, with an "X" for UNIX.

POSIX describes what a UNIX-like system must provide, without saying how to program it:

POSIX defines… Examples
The system calls that programs can use open a file, create a process
The shell language if, for, case, the redirections > and `
The basic commands and their options ls, cp, grep, sed, awk, find
Permissions and signals chmod 755, kill -9

Later, the organisation The Open Group (owner of the "UNIX" trademark since the 1990s) published the Single UNIX Specification. Only certified systems that pay for the licence are allowed to officially call themselves "UNIX". This is the case for macOS (certified since version 10.5, in 2007), AIX and Solaris.

Linux is not UNIX-certified (certification is expensive and brings nothing to a free project), but it follows POSIX very closely. That is why Linux is called a "UNIX-like" system.

getconf _POSIX_VERSION    # which version of POSIX? e.g. 200809 = 2008 edition

Practical advice for your whole career: when you write a script that must run everywhere, write it in POSIX sh rather than with tricks specific to bash. On Debian and Ubuntu, /bin/sh is in fact dash, a minimal POSIX shell: a "bash" script run with sh may fail there.

6. The lawsuit that slowed BSD down (1992-1994)

At the end of the 1980s, the Berkeley developers set out to replace all the original AT&T code in BSD, so that it could be distributed freely. In 1991, Net/2 was released, almost entirely free. Projects were derived from it: 386BSD (1992), for PCs, and the company BSDi, which sold a commercial version.

In 1992, USL (Unix System Laboratories, AT&T's UNIX subsidiary) sued BSDi, then the University of California: BSD supposedly still contained AT&T code and trade secrets. For two years, nobody knew whether BSD could continue to be distributed. Companies hesitated to use it.

In 1993, Novell bought USL. The lawsuit ended with a settlement in early 1994: a few files were removed or modified, and Berkeley released 4.4BSD-Lite (1994), free of any AT&T code.

But during those two years of uncertainty, another system, free and with no legal risk at all, advanced very quickly: Linux. Linus Torvalds said it himself: if a free BSD for PCs had been available at that time, he might never have written his kernel.

7. The descendants of BSD today

Net/2 (1991) and 386BSD (1992), then 4.4BSD-Lite (1994)
 ├── FreeBSD (1993) ── servers, Netflix (video streaming), base of the system of recent PlayStations
 ├── NetBSD  (1993) ── "runs on anything", very portable
 │     └── OpenBSD (1995) ── security first; created OpenSSH (1999)
 └── NeXTSTEP / Darwin ── base of macOS and iOS (with the Mach microkernel)

FreeBSD and NetBSD were born in 1993 from 386BSD; after the 1994 settlement, they were rebuilt on 4.4BSD-Lite.

So you may already be using BSD without knowing it: macOS and iOS descend from it, and OpenSSH, which you will use on every Linux server, is developed by the OpenBSD team.

8. The lesson timeline

1974  Berkeley receives UNIX
1978  First BSD distribution (Bill Joy)
1982  Sun founded; AT&T break-up agreement; System III
1983  System V (AT&T); 4.2BSD brings TCP/IP
1984  AT&T broken up: UNIX becomes a commercial product
1986  AIX (IBM)
1988  POSIX (IEEE 1003.1); OSF created; start of the "UNIX wars"
1991  Net/2; Linus Torvalds starts Linux
1992  USL lawsuit against BSDi; Solaris 2 (System V R4 base)
1993  FreeBSD, NetBSD; Novell buys USL
1994  Settlement: 4.4BSD-Lite, free of AT&T code

Common pitfalls

  • Believing that a command seen on macOS will work the same way on Linux. Both follow POSIX, but the options often differ: sed -i is written sed -i '' on macOS (BSD tools), sed -i on Linux (GNU tools).
  • Confusing "UNIX" and "UNIX-like". "UNIX" is a trademark reserved for certified systems. Linux and FreeBSD are "UNIX-like".
  • Writing #!/bin/sh at the top of a script full of bash syntax. On Debian and Ubuntu, sh is dash: arrays or [[ … ]] fail there. Write #!/bin/bash if you use bash.
  • Thinking that /etc/init.d/ scripts are still used. They come from System V; since RHEL 7 and Debian 8, systemd starts the services.

Lab

This time, you are on Debian 12. You will check the system's POSIX heritage for yourself:

getconf _POSIX_VERSION    # the version of POSIX followed
ls -l /bin/sh             # which shell does /bin/sh point to?
ps -ef | head -n 3        # System V syntax
ps aux | head -n 3        # BSD syntax: the same result, or nearly

The tasks: save the POSIX version, find out which shell is hiding behind /bin/sh, and write down the two main branches of UNIX. Then click Check.

Quiz

  1. What are the two main branches of UNIX, and who created each of them?
  2. Name two major contributions of BSD that you will use on Linux.
  3. What is POSIX, and why is it useful to an administrator?
  4. Is Linux allowed to officially call itself "UNIX"? Why?
  5. Why did the USL lawsuit against BSDi (1992-1994) benefit Linux?
Answers
  1. BSD, created at the University of Berkeley (Bill Joy and his colleagues), and System V, the commercial version from AT&T.
  2. Any of: the TCP/IP network stack and sockets, the vi editor, the csh shell (history, job control), the Fast File System, and more recently OpenSSH (OpenBSD).
  3. A standard (IEEE, 1988) that defines the commands, the shell and the system calls of a UNIX-like system. A script or skill that conforms to POSIX works on Linux, BSD, macOS…
  4. No: "UNIX" is a trademark reserved for systems certified by The Open Group. Linux follows POSIX but is not certified: it is a "UNIX-like" system.
  5. For two years, the legal future of BSD was uncertain; developers and companies turned to Linux, free and with no legal risk.

Reference sources: M. K. McKusick (Open Sources, 1999), POSIX and The Open Group standards documents, Debian Reference (free licence). Original MasterLinux text.

Chapter 3

GNU and free software

Objective

Understand where most of the tools you will use on Linux come from, what "free software" means, and learn to read the main licences (GPL, BSD, MIT) you will come across in your job.

Key points

  • In 1983, Richard Stallman announced the GNU Project: a complete UNIX-like system, entirely free.
  • In 1985, he created the Free Software Foundation (FSF). The GPL licence followed in 1989.
  • Free software guarantees four freedoms: to use, to study, to redistribute, and to modify and share your modifications.
  • The GPL is copyleft: anyone who redistributes a modified GPL program must redistribute it under the GPL. The BSD and MIT licences are permissive: you can do almost anything you want with the software.
  • In 1998, the term open source appeared: the same technical reality, with a message aimed more at businesses.

1. The problem: code becomes closed (late 1970s)

In the early years of computing, programmers naturally exchanged their code, just as researchers exchange their results. UNIX itself circulated with its source code in universities (lesson 0.1).

Around the start of the 1980s, everything changed: vendors realised that software could be sold separately from the machine. Source code became a trade secret. Users received programs that were already compiled, which they had no right to study, no right to fix and no right to share.

Richard Stallman, a programmer at the MIT Artificial Intelligence Laboratory, experienced this change as an injustice. A famous anecdote illustrates it: his laboratory's printer often jammed. On the old printer, he had modified the program to warn users of a paper jam. For the new one, the manufacturer refused to provide the source code. It was impossible to fix what was actually a simple problem.

2. 1983: the announcement of the GNU Project

On 27 September 1983, Stallman announced on Usenet (the ancestor of online forums) an extraordinary project: to write a complete operating system, compatible with UNIX, and entirely free. He called it GNU, a recursive acronym that means "GNU's Not Unix".

Why UNIX-compatible? Because UNIX is well designed (lesson 0.1), it is divided into small programs that can be rewritten one by one, and users would be able to move from one to the other without relearning everything.

Work began in January 1984, when Stallman left his job at MIT so that the university could not claim his code. In 1985, he published the GNU Manifesto, which explains his approach, and founded the Free Software Foundation (FSF) to fund the project.

3. The GNU building blocks you will use every day

Piece by piece, the GNU Project rewrote the UNIX tools. By the end of the 1980s, almost nothing was missing:

GNU tool Date Role Command
GNU Emacs 1984-1985 Text editor emacs
GCC (GNU C Compiler, today the GNU Compiler Collection) 1987 Compiler: it was used to compile Linux gcc
Bash (Bourne Again Shell, Brian Fox) 1989 The shell, compatible with Bourne's sh shell bash
coreutils (merger of fileutils, textutils, shellutils) 1990s-2000s The basic commands ls, cp, mv, cat, chmod
grep, sed, gawk, tar, gzip, find 1980s-1990s Searching, transforming, archiving grep, sed, awk, tar
glibc from 1987 The C library, used by almost every program (invisible, but everywhere)

Check for yourself:

ls --version | head -n 1      # ls (GNU coreutils) …
bash --version | head -n 1    # GNU bash, version …
grep --version | head -n 1    # grep (GNU grep) …

However, one essential piece was missing: the kernel, the program that talks directly to the hardware. The kernel planned by GNU, the Hurd (started in 1990), is very ambitious and has progressed slowly. It is still not used in production today. Linux would provide this missing piece in 1991 (lesson 0.4).

4. The four freedoms

For Stallman, "free" is about freedom, not price. His phrase: free as in "free speech", not as in "free beer". French is lucky enough to have two separate words: libre (free as in freedom) and gratuit (free of charge).

A program is free software if its user has the four freedoms (numbered from zero, as any good computer scientist would):

Freedom What you can do Concrete example
0 Use the program, for any purpose Install Linux on the 200 checkouts of a retail chain, without counting licences
1 Study how it works and adapt it (so you need the source code) Read the code of a web server to understand an error
2 Redistribute copies Give a colleague a USB installation key
3 Improve the program and share your improvements Fix a bug and send the patch to the project

Free software can be sold: Red Hat sells subscriptions to RHEL, which is made of free software. And software that costs nothing is not necessarily free software: many no-cost applications hide their code.

5. The GPL and copyleft (1989)

To guarantee these freedoms, Stallman uses… copyright law itself. The GNU General Public License (GPL) says, in essence:

You may use, study, modify and redistribute this program. But if you redistribute a modified version, you must release it under the same licence, with its source code.

This is the principle of copyleft (a play on the word copyright): freedom is passed on, and nobody can take it away along the way.

Version Year Key point
GPL v1 1989 First version
GPL v2 1991 The licence of the Linux kernel ("version 2 only")
GPL v3 2007 Protects against patents and against devices that prevent you from installing a modified version
LGPL 1991 A "lighter" version for libraries: a non-free program can use them
AGPL 2007 (v3) Extends the obligation to software used remotely, over the network (web applications)

An important point for an administrator: the GPL imposes nothing on you as long as you do not redistribute. You can modify GPL software for your company's servers without publishing anything. (The AGPL is the exception: if you offer the service to the public over the network, you must give the code to its users.)

6. 1998: the "open source" movement

In January 1998, the company behind the Netscape browser announced that it would publish its source code (which would become Mozilla, then Firefox). It was a shock: a large company was opening up its flagship product.

A few weeks later, in February 1998, a group of developers, including Eric S. Raymond and Bruce Perens, proposed a new term: open source, coined by Christine Peterson. They founded the Open Source Initiative (OSI). Their idea: to talk to businesses about quality, reliability and costs, rather than about ethics and freedom.

Free software (FSF) Open source (OSI)
Main argument A question of ethics: the user must be free A question of efficiency: open development produces better software
Emblematic figure Richard Stallman Eric S. Raymond, Bruce Perens
Reference text The four freedoms The Open Source Definition, derived from Debian's principles (DFSG)
In practice The recognised licences are almost the same

Raymond's essay, The Cathedral and the Bazaar (1997), compares two ways of developing software: in a small closed group (the cathedral) or in the open, with thousands of contributors (the bazaar, the Linux model). It gave us an often-quoted idea: with enough eyes, all bugs are easy to find.

You will often see the abbreviation FLOSS or FOSS (Free/Libre and Open Source Software), which brings the two camps together.

7. The licences you will come across

On a server, every piece of software has its licence. Here are the main families:

Licence Type What it requires Examples
GPL (v2, v3) Strong copyleft Redistribute modifications under the GPL, with the source code Linux kernel (v2), bash, GCC, coreutils
LGPL Weak copyleft Copyleft on the library, not on the program that uses it glibc
AGPL Network copyleft Also provide the code to the users of the online service Some web software (Nextcloud, for example)
BSD (2 or 3 clauses) Permissive Keep the copyright notice FreeBSD, OpenSSH (for the most part)
MIT Permissive Keep the copyright notice, nothing else X Window (X11), Node.js, jQuery
Apache 2.0 Permissive Copyright notice + a patent clause Apache httpd server, Kubernetes, Android (largely)

The example that makes the difference clear: Apple was able to reuse BSD code in macOS without publishing its modifications, because the BSD licence allows it. An Internet router manufacturer that modifies the Linux kernel, on the other hand, must provide the source code of its modifications to its customers, because the kernel is under the GPL.

On Debian, the texts of common licences are stored in a single directory, and each package describes its licence:

ls /usr/share/common-licenses/          # GPL-2, GPL-3, LGPL, BSD, Apache-2.0…
head -n 3 /usr/share/common-licenses/GPL-3
# On a full Debian system, each package has its licence file:
# /usr/share/doc/<paquet>/copyright

8. GNU/Linux or Linux?

Stallman asks people to say GNU/Linux, because the complete system owes a great deal to the GNU tools, and not only to the kernel. Debian follows this usage ("Debian GNU/Linux"). Most people, and most companies, simply say Linux.

In this course, we say "Linux" for convenience, but remember the technical truth: Linux is the kernel; a large part of the tools around it come from GNU. And some "Linux" systems use almost no GNU software: Android, for example, or the Alpine distribution (with BusyBox and the musl library).

9. The lesson timeline

1983  27 September: Stallman announces the GNU Project
1984  Work begins; Stallman leaves MIT
1985  GNU Manifesto; Free Software Foundation created
1987  GCC
1989  GPL v1; Bash
1990  Start of the Hurd, the kernel planned by GNU
1991  GPL v2; LGPL
1992  The Linux kernel moves to the GPL: the free system is complete
1997  "The Cathedral and the Bazaar" (Eric S. Raymond)
1998  Netscape opens its code; birth of the term "open source" and of the OSI
2007  GPL v3 and AGPL v3

Common misconceptions

  • "Free software means it costs nothing." No. Free software is about rights; free software can be sold (RHEL subscriptions, SUSE support…).
  • "Open source means anyone can modify the official software." No. You can modify your copy. For your patch to enter the official project, it must be accepted by its maintainers.
  • "The GPL forces you to publish the code of everything you develop on Linux." No. A program that runs on Linux is not affected. Only redistributed modifications of GPL software are.
  • "Free software is for amateurs." The largest websites in the world, banks and telecom operators run on free software.

Lab

On your Debian 12 machine, you will find GNU's signature and the installed licences:

ls --version | head -n 1
bash --version | head -n 1
ls /usr/share/common-licenses/
grep -m1 Version /usr/share/common-licenses/GPL-3

The tasks: note the GNU mention of ls and bash, list the licences provided by Debian, and find the date of the GPL 3. Then click Check.

Quiz

  1. What does the name "GNU" mean, and in what year was the project announced?
  2. Name the four freedoms of free software.
  3. What is copyleft? Which licence is its best-known example?
  4. A company modifies the code of a program under the MIT licence and sells it without publishing its modifications. Is it allowed to? And if the program were under the GPL?
  5. Which piece was the GNU Project missing in the early 1990s?
Answers
  1. GNU's Not Unix, a recursive acronym; announced on 27 September 1983 by Richard Stallman.
  2. 0 use for any purpose; 1 study and adapt (access to the source code); 2 redistribute copies; 3 improve and share the improvements.
  3. The principle that modified and redistributed versions must remain under the same free licence. The typical example is the GPL.
  4. With MIT: yes, it only needs to keep the copyright notice. With the GPL: no, if it redistributes a modified version, it must provide the source code under the GPL.
  5. The kernel (the Hurd was not ready): Linux provided it.

Reference sources: GNU Project texts (gnu.org), Open Source Definition (opensource.org), Debian Policy. Original MasterLinux text.

Chapter 4

1991 — Linus Torvalds and the Linux kernel

Objective

Learn how a student wrote, almost as a game, the kernel that now runs most of the world's servers, and understand how this kernel is developed and numbered. You will be able to read a version such as 5.14.0-427.el9.x86_64.

Key points

  • In 1991, Linus Torvalds, a student at the University of Helsinki, wrote a kernel for his PC, inspired by Minix.
  • He announced it on 25 August 1991; version 0.01 came out in September 1991.
  • In 1992, Linux moved to the GPL licence: together with the GNU tools, an entirely free system finally existed.
  • Linux 1.0 came out in March 1994. Today, thousands of developers, employed by hundreds of companies, contribute to it.
  • A new version comes out roughly every 9 to 10 weeks; some are maintained for a long time: these are the LTS versions.

1. What is the kernel?

The kernel is the central program of the system. It starts first and stops only when the machine shuts down. It is the link between programs and hardware:

   You ──► bash, ls, nginx, mysql…         (the programs)
                   │  "system calls": read, write, open, create a process
                   ▼
            ┌──────────────┐
            │ LINUX KERNEL │  memory, processes, disks, network, drivers
            └──────────────┘
                   │
                   ▼
        processor, memory, disks, network card      (the hardware)

When ls wants to read a directory, it does not talk to the disk: it asks the kernel. This is what allows a thousand programs to share the same machine without getting in each other's way.

2. Minix, the students' UNIX (1987)

At the end of the 1980s, UNIX was expensive and its code was no longer freely distributed. Professor Andrew Tanenbaum, at the Vrije Universiteit Amsterdam, wrote a small educational UNIX for his students: Minix (1987). It came with his textbook on operating systems, source code included.

Minix had two limits: its licence did not allow it to be modified and redistributed freely, and Tanenbaum wanted to keep it simple, for teaching. He refused many improvements proposed by its users.

3. 1991: a student in Helsinki

Linus Torvalds was born in 1969 in Helsinki. In January 1991, he bought a PC with an Intel 386 processor. He installed Minix on it and wanted to understand how this new processor worked. He started with a small program that displayed "A"s and "B"s alternately, to test switching from one task to another. Then a terminal emulator to connect to the university. Then a disk driver… and he realised that he was writing a kernel.

On 25 August 1991, he posted a message on the Usenet forum comp.os.minix. In it, he announced a free operating system for 386 PCs, specifying that it was "just a hobby, won't be big and professional like gnu". He asked readers what they would like to find in it.

Version 0.01 was uploaded to a University of Helsinki server (ftp.funet.fi) in September 1991. Linus wanted to call it Freax (free + freak + x); the server administrator, Ari Lemmke, put the files in a directory named linux. The name stuck.

From this very first version, Linux ran bash and gcc, two GNU tools (lesson 0.3).

4. 1992: the GPL and the dispute with Tanenbaum

Linus's first licence prohibited any commercial use. In early 1992, with version 0.12, he announced the move to the GPL (version 2). From then on, anyone could use, modify and redistribute Linux, provided they shared their modifications. With the Linux kernel and the GNU tools, a complete free system finally existed. Linus later said that choosing the GPL was one of the best decisions he ever made.

In January 1992, Tanenbaum posted a message on the same forum entitled "LINUX is obsolete". His two criticisms:

Tanenbaum's criticism History's answer
Linux is a monolithic kernel (the whole kernel in a single block), an outdated design in his view; the future belonged to microkernels Linux remained monolithic, but modular: drivers are loaded on demand (lsmod, modprobe)
Linux is too tied to the 386 processor and will not be portable Linux runs today on x86, ARM, POWER, IBM Z, RISC-V… more platforms than any other system

This courteous but lively debate has remained famous. It shows that Linux was not the "best" system on paper; it won because it was free, practical and open to contributions.

5. The kernel grows (1992-2011)

The timing was ideal: BSD was blocked by its lawsuit (lesson 0.2), GNU's Hurd was not progressing, and the Internet was arriving in universities. Hundreds of developers sent patches to Linus.

Version Date Notable new feature
0.01 September 1991 First public version
0.12 early 1992 Move to the GPL
1.0 March 1994 First version considered stable; TCP/IP networking
2.0 June 1996 Multiple processors (SMP); arrival of the mascot Tux, the penguin
2.2 January 1999 Better multiprocessor support
2.4 January 2001 USB, large amounts of memory: Linux enters the enterprise
2.6 December 2003 New development model; kernel of RHEL 5 and RHEL 6 (2.6.32)
3.0 July 2011 Number change for Linux's 20th anniversary, with no technical break

In 2005, a licensing conflict deprived the developers of the tool they used to manage the code (BitKeeper). In a few weeks, Linus wrote his own tool: Git. Today, Git is used by almost every developer in the world (GitHub, GitLab…). It is Linus's second major invention.

6. How the kernel is developed today

The Linux kernel is probably the largest collaborative project in the history of software: more than 30 million lines of code, thousands of contributors for each version, most of them paid by their employer (Intel, Google, Red Hat, AMD, Meta, Huawei, SUSE, Linaro…).

A regular cycle

 version 6.N released
   │
   ├── 2 weeks: the "merge window"
   │      new features prepared by the maintainers enter the kernel
   │
   ├── 6.(N+1)-rc1, rc2, … rc7 (sometimes rc8): one "release candidate" per week
   │      no more new features, only bug fixes
   │
   └── about 9 to 10 weeks later: version 6.(N+1) released

This gives 5 to 6 versions per year. There is no fixed date: Linus releases when the version is ready.

A pyramid of maintainers

Linus does not review everything himself. The kernel is divided into subsystems (networking, file systems, graphics drivers, USB, security…). Each has its maintainers, listed in the MAINTAINERS file of the source code. Contributors send their patches by e-mail to the right maintainer; the maintainer reviews them, tests them, then passes them up. Linus merges the work of the top-level maintainers.

Mainline, stable, longterm

On kernel.org, you will see several kinds of versions:

Category What it is Maintained by
mainline Linus's development version (e.g. 6.N-rc3) Linus Torvalds
stable The latest released version, with fixes (e.g. 6.N.4) until the next one Greg Kroah-Hartman and the stable team
longterm (LTS) Usually the last version of each year, fixed for several years Volunteer maintainers

Since 2023, new LTS versions on kernel.org are announced for about two years (compared with six years for some before), due to a lack of resources to maintain more. Projects such as the CIP (Civil Infrastructure Platform) maintain certain versions for ten years for industry.

The first number means nothing technical: Linus changes it when the second one becomes too large (3.0 in 2011, 4.0 in 2015, 5.0 in 2019, 6.0 in 2022…).

7. Distribution kernels

Enterprise distributions do not follow the latest kernel.org version. They choose a version at launch and maintain it themselves for the whole life of the product, adding security fixes and even new features (this is called "backporting").

Distribution Base kernel
RHEL 6 / CentOS 6 2.6.32
RHEL 7 / CentOS 7 3.10
RHEL 8 / Rocky 8 / Alma 8 4.18
RHEL 9 / Rocky 9 / Alma 9 5.14
RHEL 10 / Alma 10 6.12
Debian 12 "bookworm" 6.1
Debian 13 "trixie" 6.12
Ubuntu 22.04 LTS 5.15
Ubuntu 24.04 LTS 6.8

How to read a Red Hat kernel version:

uname -r
# 5.14.0-427.13.1.el9_4.x86_64
#   5.14.0        → base version chosen by Red Hat (kernel.org 5.14)
#   427.13.1      → Red Hat build number: hundreds of patches added
#   el9_4         → "Enterprise Linux 9", update 9.4
#   x86_64        → architecture (64-bit PC)

A RHEL 9 5.14 kernel from 2025 is therefore very different from the kernel.org 5.14 kernel of 2021: it contains years of fixes.

8. The lesson timeline

1987  Minix (Andrew Tanenbaum)
1991  January: Linus buys his 386 PC — 25 August: the announcement — September: Linux 0.01
1992  Linux moves to the GPL (0.12); "LINUX is obsolete" debate
1994  Linux 1.0 (March)
1996  Linux 2.0; the Tux mascot
2003  Linux 2.6
2005  Linus creates Git
2007  Creation of the Linux Foundation (which employs Linus)
2011  Linux 3.0, for the 20th anniversary
2022  Linux 6.0; first Rust language components in the kernel (6.1)

Common pitfalls

  • Confusing the kernel version with the distribution version. uname -r gives the kernel; /etc/os-release gives the distribution. Rocky 9.4 and Rocky 9.5 have the same 5.14 base kernel.
  • Believing that an old kernel number means "not up to date". The 4.18 kernel of RHEL 8 still receives security fixes. What matters is the package date, not the base number.
  • Installing "the latest kernel from kernel.org" on an enterprise server. You lose vendor support, and third-party modules (drivers, antivirus) may stop working.
  • In a container, believing that the kernel is the distribution's kernel. A container shares the kernel of the host machine. You will see this in the lab.

Lab

You are on an Ubuntu 24.04 machine. Query the kernel:

uname -r                           # the kernel version
uname -r | cut -d. -f1             # only the first number
cat /proc/sys/kernel/ostype        # where uname gets its answers
cat /proc/sys/kernel/osrelease
grep PRETTY_NAME /etc/os-release   # the distribution

Surprise: your lab is a container. The distribution is indeed Ubuntu 24.04, but the kernel displayed is not Ubuntu's 6.8: it is the kernel of the machine hosting the lab. A container borrows its host's kernel; only the programs and files are Ubuntu's. Complete the tasks, then click Check.

Quiz

  1. Which system was Linus Torvalds using before writing Linux, and who had created it?
  2. Why was the move to the GPL in 1992 so important?
  3. What is a "merge window" in kernel development?
  4. What is the difference between a stable version and a longterm (LTS) version on kernel.org?
  5. A RHEL 9 server displays 5.14.0-503.el9.x86_64. Why is it not "behind" even though kernel.org offers 6.x versions?
Answers
  1. Minix, a small educational UNIX created by Professor Andrew Tanenbaum (1987).
  2. The GPL lets everyone use, modify and redistribute Linux while requiring them to share improvements; it attracted contributors and made it possible to combine the kernel with the GNU tools to form an entirely free system.
  3. The two weeks following a release, during which new features enter the kernel; then come the -rc versions, reserved for fixes.
  4. A stable version is fixed only until the next one comes out; a longterm version is fixed for several years.
  5. Red Hat chose the 5.14 base and maintains it itself, adding security fixes and new features (the number 503 shows this). The base number does not tell you whether the kernel is up to date.

Reference sources: Linux kernel documentation (docs.kernel.org, GPL-2.0), kernel.org, "Just for Fun" by L. Torvalds. Original MasterLinux text.

Chapter 5

Linux distributions — Slackware, Debian, Red Hat, Ubuntu and the others

Objective

Understand what a distribution is, where the major families you will meet in companies come from, and why the command to install software changes from one server to another.

Key points

  • The kernel alone is not enough: a distribution brings together the kernel, the GNU tools, thousands of programs, an installer and a package manager.
  • The ancestors still alive were born in 1993: Slackware and Debian. Red Hat and SUSE followed in 1994.
  • Two major families dominate servers: Debian (.deb packages, apt) and Red Hat (.rpm packages, yum/dnf).
  • Ubuntu (2004) derives from Debian; Fedora (2003) prepares the future of RHEL.
  • A distribution publishes either fixed releases (Debian, RHEL, Ubuntu LTS) or continuously (rolling release: Arch).

1. Why distributions?

In 1992, installing Linux was an adventure: you had to get the kernel from a server, compile dozens of GNU programs yourself, copy them onto floppy disks, and hope that everything worked together.

A distribution does this work for you. Compare it with a kitchen:

In a kitchen In a distribution
The ingredients (flour, eggs, sugar) The Linux kernel, bash, coreutils, nginx, MariaDB…
The recipe, tested so that everything goes together The versions chosen and tested with one another
The chef and their style The distribution team, its choices, its philosophy
Regular delivery of fresh ingredients Security updates

In practice, a distribution provides:

  1. an installer (Anaconda at Red Hat, debian-installer, Subiquity at Ubuntu);
  2. a package manager to install, update and remove software;
  3. repositories: servers that host tens of thousands of ready-to-use packages;
  4. default settings: boot system, firewall, security, file locations;
  5. a release policy and a support period.

2. The pioneers (1992-1993)

The very first distributions were called MCC Interim Linux (1992, University of Manchester) and SLS (Softlanding Linux System, 1992). SLS was practical but full of bugs. Two people decided to do better.

Slackware (July 1993)

Patrick Volkerding, an American student, fixed SLS for his own use, then published the result: Slackware. It is the oldest distribution still maintained. Its philosophy: stay close to UNIX, simple, with no "magic" tools. It is very little used in companies today, but it has trained generations of administrators.

Debian (16 August 1993)

Ian Murdock, a student at Purdue University, announced on 16 August 1993 a distribution built openly by a community, and not by a single person. The name combines Debra (his girlfriend at the time) and Ian.

Debian has three characteristics that still matter:

  • It belongs to no company. Thousands of volunteers keep it alive, according to a Social Contract (1997) that promises to remain 100% free.
  • Its Debian Free Software Guidelines (DFSG) served as the basis for the definition of open source (lesson 0.3).
  • Its releases are named after characters from Toy Story: buzz (1.1, 1996), woody, squeeze, jessie, stretch, buster, bullseye (11), bookworm (12), trixie (13)…

Debian also invented a powerful package system: dpkg (mid-1990s), then APT (1998), which downloads and installs dependencies automatically. At the time, it was a small revolution.

3. The commercial distributions (1994)

Red Hat (1994)

Marc Ewing created Red Hat Linux in 1994; Bob Young partnered with him to found the company Red Hat (1995). Red Hat invented the RPM package format (Red Hat Package Manager), then made it free. Red Hat went public in August 1999, with spectacular success: Linux became serious business in the eyes of companies.

In 2003, Red Hat stopped its consumer "Red Hat Linux" to focus on RHEL (paid, for companies) and launched Fedora (free of charge, community-driven). This whole saga is detailed in lesson 0.6.

SUSE (1992-1994)

The German company S.u.S.E. (Nuremberg, founded in 1992) published its first distribution in 1994, initially based on Slackware. It became the reference in Europe, with its graphical configuration tool YaST. Today, SUSE Linux Enterprise (SLES) competes with RHEL in large companies, and openSUSE (2005) is its community version. SUSE uses RPM packages with its own tool, zypper.

4. Ubuntu: Linux for everyone (2004)

In 2004, the South African entrepreneur Mark Shuttleworth founded the company Canonical and launched Ubuntu, a word of Bantu origin that evokes humanity and sharing. The goal: a Linux that is easy, up to date, with a predictable schedule.

  • Debian base: the same .deb packages, the same apt tool.
  • One release every six months, numbered by year and month: 24.04 = April 2024.
  • One LTS release (Long Term Support) every two years, in April of even years: 20.04, 22.04, 24.04… It is supported for 5 years (and longer with the paid Ubuntu Pro subscription).
  • Animal names, in alphabetical order since 2006: jammy (22.04, Jammy Jellyfish), noble (24.04, Noble Numbat)…

Ubuntu has become the most widespread Linux in the cloud and on developers' workstations. It has derivatives of its own, such as Linux Mint (2006).

5. The other families to know

Distribution Year Packages Main idea
Gentoo 2002 Compiled from source (emerge) Optimise everything for your machine
Arch Linux 2002 pacman Simple, minimalist, rolling release; a reference wiki for all Linux systems
Alpine Linux 2005-2006 apk Tiny and secure (musl, BusyBox): widely used as a base for container images
Linux Mint 2006 apt Ubuntu made more familiar for former Windows users
Kali Linux 2013 apt (Debian base) Penetration testing tools (offensive security)
Oracle Linux 2006 rpm, dnf Oracle's rebuild of RHEL

The DistroWatch website lists hundreds of them. But in companies, remember above all two families:

DEBIAN FAMILY — .deb packages, dpkg and apt tools
  Debian (1993)
   └── Ubuntu (2004)
         └── Linux Mint, Pop!_OS…

RED HAT FAMILY — .rpm packages, rpm and yum (RHEL 5-7) / dnf (RHEL 8+) tools
  Red Hat Linux (1994-2003)
   ├── Fedora (2003) ──► CentOS Stream ──► RHEL
   └── RHEL ──► Rocky Linux, AlmaLinux, Oracle Linux (rebuilds)

OUTSIDE THESE TWO FAMILIES
  SUSE (rpm + zypper), Arch (pacman), Alpine (apk), Gentoo (emerge)

6. The package manager: the heart of a distribution

A package is an archive that contains a piece of software and information: its version, its description, and above all its dependencies (the other packages it needs).

There are two levels of tools in each family:

Level Role Debian family Red Hat family
Low level Install a package file already present, query the local database dpkg rpm
High level Search the repositories, download, resolve dependencies, update the whole system apt (and apt-get) yum (RHEL 5 to 7), dnf (RHEL 8 and later) 
# Debian / Ubuntu family
apt update && apt install nginx     # install nginx and its dependencies
dpkg -l | grep nginx                # is it installed?

# Red Hat family
dnf install nginx                   # (yum install nginx on RHEL 7)
rpm -qa | grep nginx                # is it installed?

Dependencies are the whole point: you ask for nginx, and the tool also installs the necessary libraries, in versions tested together. Without a package manager, you would spend hours looking for them one by one.

More recently, universal formats have appeared, independent of the distribution: Snap (Canonical, 2016), Flatpak (mainly for desktop applications) and, on the server side, containers (Docker in 2013, Podman). They do not replace classic packages: they come in addition to them.

7. Fixed releases or continuous publication

Model Principle Examples For whom?
Fixed releases, long support One major release every 2-3 years, then only fixes for years RHEL, Rocky, Alma (10 years), Debian (≈ 5 years with LTS support), Ubuntu LTS Production servers: stability above all
Fixed releases, short cycle One release about every 6 months Fedora, non-LTS Ubuntu Workstations, discovering new features
Continuous publication (rolling) No releases: you update all the time Arch, openSUSE Tumbleweed Enthusiasts, personal computers

A bank or online shop server wants nothing to change unless someone has decided it: it runs on a fixed release with long support. That is why this course teaches Red Hat (RHEL and its clones), Debian and Ubuntu LTS.

8. How do you know which distribution you are on?

It is the first habit to develop when you arrive on an unknown server:

cat /etc/os-release          # standard on all modern distributions
#   ID=debian, ID=ubuntu, ID="rocky"…     → the distribution
#   ID_LIKE="rhel centos fedora"           → the family
#   VERSION_ID="12"                        → the version
cat /etc/debian_version      # Debian family only
cat /etc/redhat-release      # Red Hat family only

The /etc/os-release file was standardised around 2012 with the arrival of systemd. On very old servers (CentOS 6, for example), it may be missing: in that case, use /etc/redhat-release.

9. The lesson timeline

1992  MCC Interim Linux, SLS — S.u.S.E. founded in Germany
1993  Slackware (July) — Debian (16 August)
1994  Red Hat Linux — first SuSE distribution
1997  Debian Social Contract
1998  APT (Debian) — Mandrake
1999  Red Hat goes public
2002  Arch Linux, Gentoo
2003  Fedora Core 1; RHEL becomes Red Hat's flagship product
2004  Ubuntu 4.10 (October); CentOS
2005  openSUSE; Alpine (2005-2006)
2013  Docker: containers spread
2023  Debian 12 "bookworm"
2024  Ubuntu 24.04 LTS "noble"
2025  Debian 13 "trixie"; RHEL 10

Common pitfalls

  • Copying an installation command found on the Internet without checking the family. apt install does not exist on Rocky; dnf does not exist on Debian. Look at /etc/os-release first.
  • Installing a .deb on a Red Hat distribution (or an .rpm on Debian). It is not the same format, nor the same libraries. Look for the package for your family.
  • Mixing the repositories of two releases (for example adding Debian 13 packages to a Debian 12 system). It is the best way to break a system: the dependencies no longer match.
  • Thinking that Ubuntu and Debian are identical. They share apt and .deb, but their software versions, their network tools (Netplan on Ubuntu) and their default firewall differ.

Lab

You are on Debian 12. Investigate as you would on an unknown server:

cat /etc/os-release
cat /etc/debian_version
dpkg -l | head -n 10            # the first installed packages
dpkg -l | grep -c '^ii'         # how many in total?
dpkg -s coreutils               # the details of a package

The tasks: identify the distribution, note its exact version, count the installed packages and read the version of the coreutils package. The lab is deliberately minimal: a full server often has several hundred packages. Then click Check.

Quiz

  1. Name four things that a distribution adds to the Linux kernel.
  2. What are the two oldest distributions still maintained, and in what year were they born?
  3. Which family does Ubuntu belong to? What is its package format?
  4. What is the difference between rpm and dnf?
  5. You arrive on an unknown server. What is your first command to find out which distribution it runs?
Answers
  1. Any of: the GNU tools and thousands of programs, an installer, a package manager, repositories, default settings, a release and security update policy.
  2. Slackware (July 1993) and Debian (August 1993).
  3. The Debian family; .deb packages, managed with dpkg and apt.
  4. rpm is the low-level tool (installs an .rpm file, queries the local database); dnf is the high-level tool that fetches packages from the repositories and resolves dependencies.
  5. cat /etc/os-release (and, on a very old Red Hat server, cat /etc/redhat-release).

Reference sources: official history of Debian and Social Contract (debian.org), Fedora and Red Hat documentation, Ubuntu release cycle. Original MasterLinux text.

Chapter 6

The Red Hat saga — RHEL, Fedora, CentOS, Rocky and AlmaLinux

Objective

Understand the Red Hat family, the most widespread in large companies: who does what between RHEL, Fedora, CentOS, Rocky and AlmaLinux, why CentOS disappeared, and what changes from one version of RHEL to the next. This is the common thread of the Red Hat track of this course.

Key points

  • RHEL (Red Hat Enterprise Linux) is the enterprise version, sold by subscription (support + updates), with 10 years of support per version.
  • Fedora (2003) is the community distribution where new features are prepared; CentOS Stream is the preview of the next RHEL update.
  • CentOS (2004) was a free rebuild of RHEL. At the end of 2020, Red Hat ended CentOS Linux 8 early (end of 2021); CentOS 7 ended on 30 June 2024.
  • Rocky Linux and AlmaLinux (2021) took over: free of charge and compatible with RHEL.
  • Each version of RHEL brings a major change: systemd in 7, dnf in 8, OpenSSL 3 in 9, image mode in 10…

1. From Red Hat Linux to RHEL (1994-2003)

In lesson 0.5 you saw the birth of Red Hat Linux in 1994. For nearly ten years, it was a single distribution, both free of charge (downloadable) and sold in a box with a manual and support. It went up to Red Hat Linux 9 (2003).

The problem: a new version came out about every six months. A bank cannot reinstall its servers twice a year. So in 2002 Red Hat launched Red Hat Linux Advanced Server 2.1, soon renamed Red Hat Enterprise Linux: a stable version, certified with major software (Oracle, SAP…) and with manufacturers' hardware (Dell, HP, IBM), and maintained for years.

In 2003, Red Hat permanently separated its two audiences:

For… Distribution Price Pace
Companies RHEL Paid subscription One major version every 3 to 5 years, supported for 10 years
The community, enthusiasts Fedora (Fedora Core 1, November 2003; the word "Core" was dropped in 2007) Free of charge One version about every 6 months, supported for about 13 months

Beware of a common misconception: the RHEL subscription does not pay for a "licence" in the Windows sense. The code is free software (GPL and others). The customer pays for support, updates, certifications and access to Red Hat's tools.

2. CentOS: RHEL for free (2004-2014)

Because RHEL is made of free software, Red Hat publishes its source code. Volunteers then had a simple idea: rebuild this code, remove Red Hat's logos and trademarks, and distribute the result free of charge.

This was CentOS (Community ENTerprise Operating System), whose first versions came out in 2004. The result was identical to RHEL "bug for bug", but without official support. Other projects did the same thing, such as Scientific Linux (Fermilab and CERN laboratories, discontinued with version 7).

CentOS quickly became the favourite free server:

  • of web hosting companies (the very widespread cPanel hosting panel long required CentOS or RHEL);
  • of startups, which wanted RHEL without paying;
  • of administrators, who trained on CentOS to pass the RHCSA exam;
  • of companies, for their test servers (with RHEL in production).

In January 2014, Red Hat hired the CentOS team and took the project under its wing. To many, CentOS then seemed secure for a long time.

3. CentOS Stream and the shock of December 2020

In September 2019, with CentOS 8, Red Hat also launched CentOS Stream: a version that moves forward continuously, just ahead of the next RHEL update.

BEFORE (2004-2020): CentOS follows RHEL

  Fedora ──► RHEL 8.3 ──► (a few weeks) ──► CentOS 8.3
                     source code published, then rebuilt

AFTER (2021 →): CentOS Stream comes before RHEL

  Fedora ──► CentOS Stream 9 ──► RHEL 9.x
              (what will go into the next minor RHEL update)

On 8 December 2020, Red Hat announced that CentOS Linux 8, promised until 2029, would end on 31 December 2021. From then on, the CentOS project would focus on CentOS Stream.

For users, it was a shock:

  • thousands of companies had installed CentOS 8 believing they had eight years ahead of them;
  • CentOS Stream does not suit everyone: it receives changes before RHEL, so it is no longer an exact clone of an already validated RHEL version.

CentOS Linux 7, for its part, kept its original end date: 30 June 2024. Many CentOS 7 servers were still running on that date, and some still are: you may come across them.

4. Rocky Linux and AlmaLinux (2021)

The community reacted within days. Two projects were born to take over the role of CentOS: a free, compatible rebuild of RHEL.

Rocky Linux AlmaLinux
Announcement December 2020 December 2020
Founder Gregory Kurtzer, co-founder of the CentOS project CloudLinux, a company that already made distributions for hosting providers
Origin of the name A tribute to Rocky McGaugh, another co-founder of CentOS, who had passed away "Alma" means "soul" in Spanish and Portuguese
First stable version Rocky 8.4, June 2021 AlmaLinux 8.3, March 2021
Organisation Rocky Enterprise Software Foundation AlmaLinux OS Foundation (non-profit organisation)

To move from CentOS 8 to one of them, each project provides an in-place migration script (migrate2rocky, almalinux-deploy), with no reinstallation. Other alternatives exist: Oracle Linux (a RHEL rebuild since 2006), CentOS Stream itself, or RHEL directly.

Red Hat also relaxed its offer: since 2021, the free Red Hat Developer Subscription for Individuals allows an individual to use RHEL on a small number of machines (up to 16).

June 2023: a new twist

In June 2023, Red Hat announced that CentOS Stream would become the only public source of RHEL-related code: the exact source code of each RHEL version would no longer be freely published, and would be reserved for customers.

The rebuilds had to adapt:

  • AlmaLinux chose to be binary-compatible (ABI) with RHEL, rather than a "bug for bug" copy: software designed for RHEL works on it, but there may be small differences;
  • Rocky Linux continues to aim for identity with RHEL, obtaining the code through other legal means;
  • CIQ (the company behind Rocky), Oracle and SUSE created OpenELA in August 2023, an association that publishes RHEL-compatible source code.

For you, as an administrator, the consequence is simple: what you learn on Rocky or Alma applies to RHEL, and vice versa. The commands, the configuration files and the RHCSA exam are the same.

5. The family today

             new features, every 6 months
  FEDORA  ─────────────────────────────────►   (the laboratory)
     │  about every 3 years, Fedora serves as the base for
     ▼
  CENTOS STREAM N  ───────────────────────►   (the next RHEL update, ahead of time)
     │
     ▼
  RHEL N.0, N.1, N.2…                          (the product, 10 years of support)
     │  compatible code
     ▼
  ROCKY LINUX N.x   ALMALINUX N.x   ORACLE LINUX N.x   (free of charge, compatible)

In 2018, IBM announced the acquisition of Red Hat for about 34 billion dollars, completed in July 2019. Red Hat remains an autonomous division, which continues to develop RHEL and to contribute massively to Linux, systemd, GNOME, Podman…

6. The life cycle of a RHEL version

Each major version of RHEL has a lifetime of about ten years, in two main phases, plus a paid option:

Phase Approximate duration What you receive
Full Support ≈ 5 years Fixes, security, new hardware, small new features with each minor version (9.1, 9.2…)
Maintenance ≈ 5 more years Security fixes and important bug fixes only
Extended support (ELS, paid extra) a few more years Critical fixes for those who could not migrate

Minor versions come out about every six months: 9.0, 9.1, 9.2… Rocky and Alma follow the same schedule, with a slight delay.

By version

Version Release Kernel What changes for the administrator End of maintenance
RHEL 6 November 2010 2.6.32 Boot by Upstart but services managed the old way: service httpd start, chkconfig httpd on; ext4 by default; iptables firewall; yum November 2020
RHEL 7 June 2014 3.10 Arrival of systemd (systemctl); firewalld firewall; XFS by default; GRUB 2; 64-bit only; MariaDB replaces MySQL; chronyd for time June 2024
RHEL 8 May 2019 4.18 dnf replaces yum (the yum command remains as an alias); BaseOS and AppStream repositories with "modules"; nftables under firewalld; Podman instead of Docker; Python 3 (no python command by default); Cockpit web console; system-wide cryptographic policies May 2029
RHEL 9 May 2022 5.14 Built from CentOS Stream 9; OpenSSL 3; SSH root login with a password refused by default; SHA-1 signatures deprecated; old ifcfg network files deprecated in favour of NetworkManager files; x86-64 processors of level v2 required May 2032
RHEL 10 May 2025 6.12 Image mode (the system deployed as a container image, with bootc); first post-quantum cryptography algorithms; Lightspeed command-line assistant; Xorg graphical server removed (Wayland only); Valkey replaces Redis; x86-64 processors of level v3 required planned for 2035

This table is the reason for the Red Hat track of this course: every important lesson there has its "By version" section, because in companies you will still come across RHEL or CentOS 6 and 7 servers alongside RHEL 9 and 10.

# On RHEL 6 / CentOS 6                        # On RHEL 7 and later
service sshd restart                          systemctl restart sshd
chkconfig sshd on                             systemctl enable sshd
iptables -L                                   firewall-cmd --list-all
yum install httpd                             dnf install httpd   (RHEL 8+)

7. The lesson timeline

1994  Red Hat Linux
1999  Red Hat goes public
2002  Red Hat Linux Advanced Server 2.1 → soon "RHEL"
2003  Red Hat Linux 9, the last one; RHEL 3; Fedora Core 1
2004  First CentOS versions
2007  RHEL 5
2010  RHEL 6
2014  Red Hat hires the CentOS team (January); RHEL 7 (June)
2019  RHEL 8 (May); IBM completes the acquisition (July); CentOS Stream (September)
2020  8 December: early end of CentOS Linux 8 announced
2021  AlmaLinux 8.3 (March), Rocky Linux 8.4 (June); end of CentOS Linux 8 (31 December)
2022  RHEL 9 (May), followed by Rocky 9 and AlmaLinux 9
2023  RHEL source code reserved for customers (June); OpenELA (August)
2024  End of CentOS Linux 7 (30 June)
2025  RHEL 10 (May), followed by AlmaLinux 10 and Rocky 10

Common pitfalls

  • Saying "CentOS" when talking about a recent Red Hat server. CentOS Linux no longer exists: a CentOS 7 still in service has received no free fixes since July 2024. It is a security risk that should be reported.
  • Confusing CentOS Stream with the old CentOS. Stream is ahead of RHEL, not a copy of a validated RHEL version.
  • Believing that Rocky and Alma are "less professional". They are used in production by large organisations. The difference with RHEL lies in support and vendor certifications, not in the commands.
  • Applying a RHEL 6 procedure on RHEL 9 (or the reverse). Always check the version (cat /etc/redhat-release) before following a tutorial.

Lab

You are on Rocky Linux 9. Check its kinship with RHEL:

grep -E '^(ID|ID_LIKE|VERSION_ID)=' /etc/os-release
cat /etc/redhat-release       # present on the whole Red Hat family
rpm -qa | wc -l               # number of installed RPM packages
rpm -q rocky-release          # the package that "signs" the distribution

The tasks: note the family declared by ID_LIKE, the line of /etc/redhat-release, the number of packages and the exact version of rocky-release. Then click Check.

Quiz

  1. What is the difference between Fedora and RHEL?
  2. What was CentOS, and why was it so popular?
  3. What did Red Hat announce on 8 December 2020?
  4. Where does CentOS Stream stand in relation to RHEL?
  5. In which version of RHEL did systemctl and firewalld appear? And dnf?
Answers
  1. Fedora is community-driven and free of charge, with a version about every 6 months: it is the laboratory for new features. RHEL is the enterprise version, sold by subscription, stable and supported for 10 years.
  2. A free rebuild of the RHEL source code, without the Red Hat trademarks: a free RHEL, identical "bug for bug", without official support. Hosting companies, startups and administrators in training loved it.
  3. That CentOS Linux 8, planned until 2029, would end on 31 December 2021, with the project focusing on CentOS Stream.
  4. Ahead of RHEL: it contains what will go into the next minor RHEL update.
  5. systemctl and firewalld: RHEL 7 (2014). dnf: RHEL 8 (2019).

Reference sources: Red Hat Enterprise Linux life cycle and release notes (Red Hat documentation, CC BY-SA 3.0), official announcements from CentOS, Rocky Linux and AlmaLinux. Original MasterLinux text.

Chapter 7

Linux today and your career

Objective

See where Linux is used today, which jobs it opens up, which certifications count on a CV, and choose your path in this course: the Red Hat family or the Debian family.

Key points

  • Linux runs the vast majority of web and cloud servers, 100% of the 500 most powerful supercomputers (since 2017), Android and a host of embedded devices.
  • The main jobs: system administrator, DevOps engineer, SRE, cloud engineer, not forgetting security and embedded systems.
  • The reference certifications: RHCSA then RHCE (Red Hat), LPIC-1 (LPI, vendor-neutral), LFCS (Linux Foundation).
  • Choose one family to start with: Red Hat (large companies, banks, telecoms, public administrations) or Debian/Ubuntu (hosting providers, startups, cloud).
  • About four fifths of what you will learn is common to both families: the core track of this course.

1. Where does Linux run?

Linux is the most widely used system in the world… even though few people see it. It is everywhere there is no Windows or macOS screen in front of you.

Field Position of Linux Examples
Web servers Vast majority Most of the websites you visit run on Linux, often with nginx or Apache
Cloud Vast majority The virtual machines of AWS, Google Cloud, Azure; Microsoft has stated that more than half of Azure's virtual machines run Linux
Containers Almost exclusive Docker, Podman and Kubernetes rely on Linux kernel features (namespaces, cgroups)
Supercomputers 100% of the TOP500 list since November 2017 Weather forecasting, medical research, nuclear simulation
Phones Android uses the Linux kernel More than 3 billion active Android devices according to Google (2021)
Embedded Very present Internet boxes, routers, smart TVs, cars, cash registers, ATMs, medical devices
Space Present The Ingenuity helicopter, on Mars in 2021, ran on Linux
Desktops Minority (a few percent) Developers, ChromeOS (based on Linux), the Steam Deck console

Remember this: when you check your bank account on your phone, it is very likely that Linux is present at every step: in the phone (Android), in the Internet box or the cell tower, in the bank's web server, and in its database.

2. Why Linux won

Going back over the six previous lessons, Linux's success can be summed up in five reasons:

  1. It is free (lesson 0.3): no licence to pay per machine. For a hosting provider with 10,000 servers, that changes everything.
  2. It is open: any company can adapt it to its hardware (Android, Internet boxes, supercomputers).
  3. It follows the ideas of UNIX (lessons 0.1 and 0.2): solid, proven, and known to generations of engineers.
  4. It is developed by everyone (lesson 0.4): competitors (Intel and AMD, Google and Microsoft) improve the same kernel together.
  5. There are distributions for every need (lessons 0.5 and 0.6): RHEL for banking, Ubuntu for the cloud, Alpine for containers.

3. The jobs

Knowing how to administer Linux opens up several jobs, which often overlap. Job titles vary from one company to another.

Job What you do every day Key Linux skills
Linux system administrator Install, configure, update and monitor servers; manage accounts, disks, backups, incidents This whole course: services, networking, storage, security, logs
DevOps engineer Automate application deployment; continuous integration pipelines; containers Shell scripts, Git, Ansible, Docker/Podman, Kubernetes
SRE (Site Reliability Engineer) Guarantee the reliability of a service at large scale; measure, alert, reduce outages. The role was formalised at Google in the early 2000s Monitoring, performance, automation, programming
Cloud engineer Design and manage infrastructure on AWS, Azure or Google Cloud Linux + networking + the provider's tools, Terraform
Security engineer Harden servers, analyse logs, respond to incidents SELinux/AppArmor, firewalls, SSH, auditing, logs
Embedded systems engineer Run Linux on a device (Internet box, car, payment terminal) Kernel, compilation, boot, drivers

A common career path: support technician or junior administrator → experienced system administrator → DevOps, SRE or cloud. In every case, the foundation is the same: being comfortable on the command line and understanding what is happening in a Linux system. That is exactly what this course builds.

In Morocco as elsewhere, there are many employers for Linux profiles: banks and insurance companies, telecom operators, public administrations, hosting providers, IT service companies and outsourcing centres (offshoring), startups.

4. The certifications

A certification proves to a recruiter that you know how to do things, not just that you have read about them. The main ones:

Certification Organisation Format Level For whom?
RHCSA (exam EX200) Red Hat 100% hands-on: you administer real RHEL machines for a few hours Administrator The reference in the Red Hat family; in high demand in companies
RHCE (EX294) Red Hat Hands-on, focused on automation with Ansible (since RHEL 8); requires the RHCSA Experienced administrator After the RHCSA, to aim for DevOps
RHCA Red Hat Several hands-on specialist exams Expert Architects
Linux Essentials LPI Multiple-choice Beginner First validation of the basics
LPIC-1 (exams 101 and 102) LPI (Linux Professional Institute) Multiple-choice and fill-in-the-blank questions Junior administrator Vendor-neutral: covers Red Hat and Debian
LPIC-2, LPIC-3 LPI Questions Experienced, expert Continuation of the LPI path
LFCS The Linux Foundation Hands-on, online, proctored Administrator A hands-on, vendor-neutral alternative to the RHCSA
CompTIA Linux+ CompTIA Questions and scenario-based tasks Junior Recognised mainly in North America

Hands-on exams (RHCSA, RHCE, LFCS) are often the most valued by recruiters: you cannot pass them by learning by heart. Certifications have a limited validity period (from two to five years depending on the organisation): check the current rules on the official website before registering, as they change regularly.

After a first Linux certification, many administrators move on to Kubernetes (CKA, from the Linux Foundation) or to the cloud providers' certifications (AWS, Azure, Google Cloud), all of which assume a good command of Linux.

5. Choosing your path: Red Hat family or Debian family?

This course offers a core track, then three tracks: Red Hat, Debian and Ubuntu. Here is how to choose.

If you are aiming for… Start with Why
A bank, an insurance company, a telecom operator, a public administration, a large company Red Hat (RHEL, Rocky, Alma) These organisations buy RHEL support; the RHCSA is often requested there
The RHCSA certification Red Hat The exam is taken on RHEL; Rocky and Alma are perfect for practice
A hosting provider, a startup, a web agency Debian or Ubuntu These are the most common distributions there
Cloud and DevOps Ubuntu, then Red Hat Ubuntu is very widespread in the cloud; client companies, on the other hand, often have RHEL
A vendor-neutral certification (LPIC-1, LFCS) The core track, then both families These exams cover both worlds

The main differences you will come across:

Red Hat family Debian / Ubuntu family
Packages .rpm — dnf (yum before RHEL 8) .deb — apt
Usual firewall firewalld nftables (Debian), ufw (Ubuntu)
Mandatory access control SELinux AppArmor
Administrators group wheel sudo
Network configuration NetworkManager (nmcli) /etc/network/interfaces (Debian), Netplan (Ubuntu)
Support for a version 10 years ≈ 5 years (Debian with LTS, Ubuntu LTS; longer with Ubuntu Pro)

The right advice: do not look for the "best" Linux. Learn one family thoroughly, then the differences of the other. An administrator who masters Rocky adapts to Ubuntu in a few days, and vice versa: the basic commands, systemd, SSH, permissions and the shell are the same.

6. How to progress

  1. Practise every day, even for twenty minutes: system administration is learned with your fingers, not only with your eyes. Every lesson in this course has a lab for this.
  2. Break things and fix them: in a lab, nothing is serious. Delete a file, stop a service, and work out how to undo it.
  3. Read error messages all the way to the end: nine times out of ten, the solution is written there.
  4. Learn to find documentation: man, --help, the official documentation of your distribution (core track, lesson on documentation).
  5. Keep a notebook of your commands and solutions: it is the start of your own procedures, and a real asset in job interviews.
  6. Aim for a certification: an exam date in your calendar is the best motivation.

7. The History module timeline

1964  Multics
1969  UNIX (Thompson, Ritchie, Bell Labs)
1973  UNIX rewritten in C; pipes
1978  BSD (Berkeley)
1983  System V (AT&T) — GNU Project (Stallman)
1985  Free Software Foundation
1988  POSIX
1989  GPL
1991  Linux 0.01 (Linus Torvalds)
1993  Slackware, Debian
1994  Linux 1.0; Red Hat Linux; SuSE
1998  "Open source"
2003  RHEL and Fedora
2004  Ubuntu; CentOS
2008  First Android phone
2017  Linux on 100% of TOP500 supercomputers
2019  IBM acquires Red Hat
2021  Rocky Linux and AlmaLinux
2024  End of CentOS 7
2025  RHEL 10, Debian 13
…     and what comes next is yours to write

Common misconceptions

  • "Linux is for geeks, not for companies." It is the opposite: Linux is above all an enterprise system. It is on personal computers that it remains a minority.
  • "With the cloud, Linux administrators are no longer needed." The cloud moves the work, it does not remove it: cloud virtual machines and containers are Linux systems that must be configured, secured and troubleshot.
  • "You need to know how to program to administer Linux." Not to start with. Shell scripts come naturally with practice; a language such as Python becomes useful later, for DevOps or SRE.
  • "A certification is enough to get hired." It opens doors, but the technical interview checks your practice: labs count as much as the diploma.

Lab

The last lab of the module, on Ubuntu 24.04:

grep '^VERSION_CODENAME=' /etc/os-release   # the code name of this version
uname -m                                    # the processor architecture

Note the code name and the architecture, then commit yourself: write the family you choose (redhat or debian) and the first certification you are aiming for. Nobody but you will read these files; it is a promise to yourself. Then click Check, and head to the core track.

Quiz

  1. Name four fields where Linux is used, apart from desktop computers.
  2. What is the main difference between a system administrator and an SRE?
  3. Why is the RHCSA particularly valued by recruiters?
  4. You want to work in a bank that uses RHEL. Which family do you learn first, and on which free distribution can you practise?
  5. Name three differences between the Red Hat family and the Debian family.
Answers
  1. Any of: web servers, cloud, containers, supercomputers, Android phones, Internet boxes and routers, TVs, cars, medical devices, space.
  2. The administrator keeps the servers running (installation, configuration, maintenance); the SRE focuses on the measured reliability of a service at large scale, with a lot of automation and monitoring. The two jobs overlap widely.
  3. Because the exam is 100% hands-on: you administer real machines, so you cannot pass by learning answers by heart.
  4. The Red Hat family; practise on Rocky Linux or AlmaLinux (or RHEL with the free developer subscription).
  5. Any of: .rpm/dnf packages versus .deb/apt; SELinux versus AppArmor; the wheel group versus sudo; firewalld versus nftables/ufw; NetworkManager versus interfaces/Netplan; support period (10 years versus about 5 years).

Reference sources: TOP500 (top500.org), official certification pages of Red Hat, LPI and the Linux Foundation. Original MasterLinux text.

Time to practise

The fundamentals start with your first terminal, for free.

Discover the fundamentals

https://masterlinux.online