Discovery
| Parameter | Value |
|---|---|
| Announcement date | 25 August 1991 (Usenet post on comp.os.minix) |
| First public release | 17 September 1991 (version 0.01 on ftp.funet.fi) |
| Author | Linus Benedict Torvalds (born 1969, Helsinki) |
| Motivation | Frustration with MINIX's limitations and UNIX's cost |
| Development machine | Intel 386 DX-33 MHz, 4 MB RAM |
| Initial licence | Ad hoc licence (no commercial use) → GPL v2 (February 1992) |
| Size of v0.01 | 10,239 lines of C, 88 files |
Technical Explanation
1. Modular monolithic architecture. Linux uses a monolithic kernel: all code — memory management, filesystems, drivers, networking — runs in kernel space (ring 0 on x86). This contrasts with microkernels (Mach, MINIX 3) where only minimal services reside in kernel space. The advantage: performance (no context-switch overhead between the kernel and user-space services), at the cost of higher risk (a driver bug can crash the entire system). Loadable kernel modules (.ko) have mitigated this risk since version 1.2 (1995).
2. The GPL v2 licence: the legal hack. Torvalds adopted Stallman's GPL v2 in February 1992. This "copyleft" licence requires that any redistributed modification remains open source. This created a virtuous circle: IBM, Red Hat, and Google contribute to the kernel because they benefit from everyone else's contributions. In 2024, corporate commits account for ~75% of development (Huawei, Intel, Red Hat, and Google leading).
3. The "bazaar" development model. Eric Raymond formalised it in 1997: "Given enough eyeballs, all bugs are shallow" (Linus's law). The kernel receives 7.8 to 8.5 accepted changes per hour. The release cycle is ~9 weeks (since the time-based model adopted in 2003). Torvalds maintains the main branch; subsystems (networking, filesystem, drivers) are delegated to ~1,700 maintainers via a hierarchical trust tree.
4. Hardware portability. Linux supports 30+ CPU architectures (x86, ARM, RISC-V, MIPS, s390, PowerPC…). The Hardware Abstraction Layer (HAL) and the driver macro-architecture allow the same kernel to be compiled for an IBM z16 mainframe server and a Raspberry Pi Zero (1 GHz, 512 MB RAM). This portability explains its dominance in embedded systems and IoT.
Why It Worked
Three converging factors: (1) timing — UNIX was expensive ($3,000–$10,000), MINIX was pedagogical only, and BSD was mired in the AT&T lawsuit (1991–1994); (2) the GPL v2, which attracted corporate contributions without allowing private appropriation; (3) the nascent Internet, which enabled a global network of developers to collaborate via mailing lists and FTP servers. Had BSD not been frozen by the USL/AT&T lawsuit, Linux might never have grown beyond a hobby.
The monolithic-but-modular approach also proved decisive. Where academic consensus favoured microkernels, Torvalds chose pragmatism — a debate with Andrew Tanenbaum in January 1992 that became legendary. Three decades later, the engineering bet paid off: loadable modules gave Linux the extensibility of a microkernel without sacrificing monolithic performance.
Causal Chain
UNIX too expensive + MINIX limited + BSD in litigation (1991) → Torvalds creates Linux 0.01 on i386 → Switches to GPL v2 (1992) → First distributions (MCC Interim Linux and SLS, 1992; Slackware, 1993, the oldest still maintained) → Server adoption (Apache + Linux, 1995) → Red Hat IPO (1999) → Android adopts the Linux kernel (2008) → Cloud (AWS, GCP) built on Linux → 100% of supercomputers → Backbone of global infrastructure
Anecdote
Torvalds originally named the project "Freax" (free + freak + the x from Unix). Ari Lemmke, the university's FTP server administrator, thought the name was ridiculous and created the directory under "linux" without asking permission. The name stuck.
Sources
References verified during the August 2026 fact-checking audit: these are the pages
against which this bulletin's claims were checked.