Even in particle/astrophysics we rely on proprietary FPGA Vendor tools, in proprietary electronics design tools (ok, kicad is seeing increasing adoption), and often proprietary embedded tools (depending on the microcontroller vendor). Not to mention proprietary graphics drivers (CUDA), CAD tools (Solidworks, Solidedge), mechanical or e&m simulation (comsol, Ansys, xfdtd, wipl-d, etc). A fair number of people use IDL or Matlab too and mathematica is pervasive among theorists. Probably nobody uses Origin
anymore. We've gone backwards on documents since everybody is using overleaf now. It's true most of the software we develop, if made public, is open source.
Yes there are still pockets of proprietary tools but is of limited usage. Most of people in the field is not going to use them (because they don't work on these things). And most of these tools are tied to the hardware and there is increasing adoption like you said for kicad and FreeCAD. There are two exceptions which are theorists love for Mathematica and CUDA for GPU programming on NVIDIA GPUs. For CUDA, it is not bad as you don't pay per usage or don't have contacts, you do this because you purchased NVIDIA GPUs. Which is the best in the world for the use case we have in particle physics experiments. I have never seen someone in the HEP community or Astrophysics using Matlab.
Regarding overleaf, it is open source and you can self-host the community edition for free or self-host professional instance and pay subscription.
Matlab users certainly exist, though they are much rarer than 15-20 years ago (before python was a suitable alternative). For signal processing, matlab still has has advantages over e.g. scipy.signal, which has not yet reimplemented all matlab functionality.
I didn't realize Overleaf was open source (or at least open core...)
And FreeCAD can't yet even effectively render many of the parts for my experiments :(
I was talking about particle physics and HEP in general as this is my field. Not many people do signal processing so it is small pocket as I said and some are using scipy just fine.
Overleaf is open source for all purposes. this does not mean you have to use all features for free selfhosted. it has to get some money. The decision by CERN to provide professional account for all CERN users on overleaf.com instead of self-hosting one has to do with minimizing unnecessary deployment and maintenance (including security) burden. I guess also they got a good discount from overleaf folks.
I'm not saying that there is complete independence now. but for most people in the field and most purposes, you can rely on open source tools and it will work and you call it a day. CERN is not the business of designing GPUs to compete with NVIDIA (Although it will be good).
