[dguest]

CERN (also in CH) made a half-effort to switch away from MS a few years ago. MS had started charging them a crazy amount of money. They got a few people working on it and even switched a few of the back end services. And actually the open source stuff worked amazingly well!

Then like a year later they doubled down on MS products (right after a new IT head came in). The IT people I spoke to had no idea why this happened but no one seemed to think going back to MS was a good idea.

Discussed more here https://news.ycombinator.com/item?id=41717607

[ahartmetz]

When I was involved in high-energy physics (doing some pretty pedestrian software stuff), CERN switched from its beloved Hypernews system to something based on MS Sharepoint. Everybody was baffled about why they would do that and hated the new system.

It seems like Hypernews was only turned off in 2021, much much later than planned, but they did do it.

[dguest]

Everything I see is moving to Discourse [1] now.

The experiment SW is a lot like the buildings and labs in every physics department I've been to: everyone is on a separate 3-5 year grant cycle and as such is contractually obliged not to push for newer infrastructure. People stick to the same software/lab until it is discontinued/condemned.

[1]: https://www.discourse.org/

[anthk]

They even published Scientific Linux, a CentOS like RHEL fork.

[dguest]

The CERN experiments and CERN IT have contributed quite a lot to open source. Part of this is necessity: when your experiment draws thousands of collaborators from hundreds of institutions and dozens of different funding agencies it's really difficult to deal with licensing fees.

Scientific Linux is discontinued, though. A few experiments went to CentOS and (when that was moved to CentOS Stream) to AlmaLinux. But practically speaking the OS the experiments are using is a RHEL-like base with almost everything important overwritten via LD_LIBRARY_PATH, PATH, etc. and pointing to a fuse-mounted file system called cvmfs

https://github.com/cvmfs/cvmfs

For better or worse this allows O(weekly) releases that change what would normally be core components of the OS.

It's kind of weird how all the interesting stuff at CERN is linux and open source, and then all the IT infrastructure is outdated MS services and Windows.

[pjmlp]

It was already like that back in 2003, the only place folks were using Linux was on the accelerator infrastructure, everyone was either using Windows or OS X as their main computer, and IT used to have a page about zero support for Linux based workstations.

I was one of the few Linux heads at the time on my group, and took part on the Scientific Linux rollout experiment.

On my group, most papers were either authored on FrameMaker or Word with a LaTeX like template.

The reason is the same as on the last agencies I have worked for since then, IT doesn't want to support Linux hardware, just like OEMs don't want to sell Linux computers (yes there are a few exceptions like System76 and TUXEDO), and most folks that really need it, get by with UNIX on macOS, or running guest VMs on Windows.

The large majority of CERN researchers aren't using Linux directly, and writing code for LHC experiments, hence why.

[dguest]

Interesting that there seems to be a shift toward Linux, if a slow one.

Within my experiment (which has a few thousand collaborators) every paper is written in LaTeX and has to build in a linux environment to be submitted for publication. All the central software work is done in one of three ways:

- ssh to a cluster running linux, develop there

- work in a linux VM on your machine (usually docker, lima, or other container management products)

- with a linux laptop or desktop

Of course the engineers and hardware developers are often using a whole Windows-dependent stack of FPGA design software, LabView, etc.