This is the English reading version of the Dutch news item CADRS overweegt een andere kernel: van OpenCASCADE naar Solidcore.
CADRS is an open-source parametric CAD program: sketches, constraints, features, assemblies and technical drawings, with a workflow you know from Onshape. It models on the OpenCASCADE kernel. In the project’s development discussion the question is now on the table whether to trade that kernel in for Solidcore.
Solidcore is a rebuild in Rust of SMLib, the kernel part of NVIDIA’s open-source USD BRep project. The first customer is sister project SolidCraft, which wants to replace its own Truck kernel with it. Solidcore’s progress page counts 261 parts: 108 are done, 89 partial, and about 70% of the lines have been ported. That page is rebuilt after every change.
Why that matters: the kernel is the part that actually makes your model. A fillet, a boolean, an offset and the export to STEP come out of it. If something goes wrong there, it is not your sketch failing but the kernel underneath. The reporter estimates that about a third of the failed operations in SolidCraft sit in the kernel. The maker of CADRS keeps it short: OCCT is somewhat slow, and a switch is in principle the conversion of one module.
Meanwhile he is not sitting still. He has spent a few days on a kernel of his own: OpenCASCADE again in Rust, intending it to compute on the graphics card instead of the processor. That kernel now has to pass the whole OpenCASCADE test set. His interim position on Solidcore: too young and too strongly interwoven with SMLib to take code from, at most the surface intersection and the point projection will be usable as a reference later.
My verdict: this is the discussion that really matters in open-source CAD. The kernel decides what your model can handle, and OCCT is a commercial organisation where doing business is a burden. Just do not let the outcome be a production plan yet: Solidcore is not finished, and a kernel that computes on the graphics card is for now in the test phase of one maker.
What a kernel does and why the choice is so decisive is on NURBS & modeling kernels. How I test this kind of switch is on Tests & benchmarks. Keep an eye on this site for updates and tests, I keep following CADRS and publish here what the kernel switch delivers.