Skip to content

The project

Where we are going, and why.

Mission

An operating system that keeps working when its parts fail: isolated, self-healing and small enough to understand and audit.

Vision

Fault resilience as the norm in servers, in classrooms and, in time, on the desktop.

Values

What we do not negotiate.

Measurable reliability

If continuous integration does not cover it, it is not guaranteed.

Small and readable

Code one person can read and understand, starting with the kernel.

Secure by design

Isolation between components, non-executable memory and ongoing audits.

Transparency

Documented decisions, the real state of every feature, no empty promises.

Respect for the legacy

We build on MINIX 3 and the work of Andrew S. Tanenbaum and his team, and we say so.

Roadmap

Step by step.

  1. Done in 2026

    • amd64 port: SMP up to 32 CPUs, hybrid UEFI/BIOS boot, a 128 TB user address space
    • MFS v4 with journaling, extended attributes and ACLs; ext4, FAT, exFAT and tmpfs
    • Memory reclaim and compression; shared file and anonymous mappings
    • POSIX: *at() calls, locks, job control, message queues, semaphores, asynchronous I/O, kqueue
    • Real-time threads: SCHED_FIFO/RR, priority inheritance and ceilings, CPU affinity, cancellation
    • Hardening: SMEP/SMAP, stack protector, FORTIFY, ASLR, bcrypt passwords; OpenSSH and TLS certificates
    • LLVM/clang 22 toolchain; TCP window scaling, SACK and timestamps
    • CI with QEMU/KVM on every pull request; the Open POSIX Test Suite
  2. Now · first release

    • Merge devel into master and carry the new name into the system
    • A usable desktop ISO: Wayland, LXQt and Qt 6
    • Release images, checksums and up-to-date release notes
  3. Next

    • Position-independent executables and RELRO; hardening for system services
    • A debugger and binary packages (pkgin)
    • Swap to disk: the tools to turn it on (the memory manager side is done)
    • Validation on reference physical hardware
    • Public resilience demo and the education track
  4. Later

    • A stable Wayland desktop
    • A wider USB stack: keyboards, mice and more
    • Live update on amd64
    • [GOALS TO DEFINE]

Updated November 2026. Each item links to the work in the repository when it lands.

Origin

From MINIX 3 to Ixolo.

MINIX was born in 1987 as a teaching system, written by Andrew S. Tanenbaum to go with his book Operating Systems: Design and Implementation. It was a microkernel from the start. MINIX 3 (2005) aimed it at reliability: device drivers moved out of the kernel into isolated processes, and a Reincarnation Server restarts them when they fail. Years of research at the Vrije Universiteit Amsterdam followed, funded by Dutch and European research grants.

Funding ended around 2016. The last stable upstream release is 3.3.0 (2014), the last release candidate 3.4.0rc6 (2017), and the last commit dates from 2018. Ixolo starts from that work and takes it to modern hardware: a port to x86-64, multiprocessor support, UEFI, new drivers and file systems, a modern LLVM toolchain, POSIX conformance work and systematic hardening.

Ixolo is an independent project maintained by David Montero. It is not affiliated with or endorsed by the Vrije Universiteit Amsterdam, the MINIX 3 project or the Stichting MINIX Research Foundation. MINIX is © Vrije Universiteit Amsterdam, under a BSD-style licence.

  1. 1987

    MINIX 1, a teaching system

  2. 1997

    MINIX 2, POSIX-oriented

  3. 2000

    Released under a BSD-style licence

  4. 2005

    MINIX 3: user-mode drivers and self-repair

  5. 2012

    MINIX 3.2: NetBSD userland and clang

  6. 2014

    MINIX 3.3: first ARM release, last stable upstream

  7. 2017

    MINIX 3.4.0rc6, last upstream release candidate

  8. 2026

    Ixolo: amd64, SMP, UEFI