Abandoning Scientific Linux Was a Mistake

In 2026, CERN expects over 2,200 industrial computers and embedded systems on its accelerator complex to run Debian 13. While Red Hat Enterprise Linux (RHEL) and AlmaLinux remain the primary supported Linux distributions across many organizations, the shift drew attention to a decision made a decade ago: CERN and Fermilab's gradual abandonment of Scientific Linux. Reflecting on this choice, it is now considered a mistake.

The abandonment of Scientific Linux was not due to technical superiority, cost-free maintenance, or control over Red Hat or IBM's decisions. The primary issue was viewing Scientific Linux primarily as duplicated engineering work. It was infrastructure designed to address a specific problem in High Energy Physics (HEP). Large experiments involved multiple laboratories, universities, computing centers, and countries.

Each site may use different versions of glibc or have varying packaging environments, leading to compatibility issues. Scientific Linux offered a stable, long-lifecycle Linux distribution that remained consistent across institutions without requiring commercial licenses for every machine.

Red Hat Enterprise Linux provided stability and a long lifecycle, while Red Hat published the source needed to build it. Scientific Linux turned this source into a community resource. Initially supported by CERN, it expanded to Fermilab and then to a broader range of users, including universities, research institutions, experiments, companies, and even the International Space Station.

When Red Hat and CentOS merged in 2014, moving to CentOS seemed like a reasonable choice. CentOS offered the same benefits as Scientific Linux, with the added advantage of being freely available due to a larger general-purpose community. CERN began transitioning its next major release from Scientific Linux to CentOS 7, while Scientific Linux 5 and 6 continued to be supported. The rationale was that spending engineering time rebuilding a Linux distribution when CentOS already provided a similar service was unnecessary.