Why replace Rap3d with jolt? As far as I know, Rap was developed using Rust and the documentation is better. I also have a good roadmap for Rap's development plan,

[WREQI]

[proggeramlug]

Thanks for asking — the rationale is in docs/physics.md under "Why Jolt (and not PhysX or Rapier)". Short version:

• Rapier is genuinely solid for rigid-body dynamics, and we agree the Rust-native story and docs are good. But it tops out there — no cloth/soft bodies, no built-in vehicle controller, no character controller with stair-stepping. Bloom needs all three.
• PhysX 5 is excellent but doesn't target WebAssembly. Picking it would have meant either dropping the web target or maintaining two physics backends.
• Jolt ships a first-party WASM build (JoltPhysics.js), is permissively licensed, faster than PhysX on rigid-body benchmarks in the public comparisons, and powers Horizon Forbidden West — so it covers our feature gaps and our web target without forcing a split backend.

This isn't a knock on Rapier — if your roadmap covers the features we needed, it's a great choice. For Bloom specifically, Jolt was the only option that hit all three of feature breadth, web support, and license.

[WREQI]

https://github.com/lo-th/phy ok. Maybe you can take a look at this, it has created an abstraction adaptation layer for a physics engine. Each of these physics engines has its own strengths. For example, PhysX provides industrial-grade simulation, although JoltPhysics is used now. It is recommended that adapting multiple physics ports can be considered as a long-term planning roadmap.