Please enable JavaScript.

Coggle requires JavaScript to display documents.

ST Techincal - Coggle Diagram

- - - - https://www.notion.so/arpss/CRDTs-for-ST-18f3631daff28063ae71d44ee0c7467e?pvs=4
- - - - Instancing: Normally, each copy of an object (like bolts in a CAD model) would need its own draw call, which is slow. With instancing, the GPU draws all copies in one go, massively reducing CPU-GPU communication. Instancing stores it ONCE in GPU memory and just applies different transformations for each instance, reducing CPU overhead.
      - Frustum culling removes objects that you can’t see at all to reduce the number of objects being processed.
      - LOD simplifies objects (reducing triangle count) that are still visible but far away.
    - - Progressive Rendering, independent of camera distance loads basic version/ wireframe. Is temporal (time-based streaming). & gradually loads more detail over time.
- - - - skipped unnecessary PBR materials, high res textures, flat or baked over dynamic lighting
- - - - const u: User = { name: "Alice" }; // ✅ Works, even without calling new User()
      - // BUT... TypeScript doesn’t enforce privacy at runtime:
        console.log((u as any).name); // 😱 ✅ Works at runtime
        
        private is just a compiler hint,
        
        At SpatialThink, we built a 3D self-authoring platform where engineers relied on TypeScript’s type system to maintain scene graph integrity. However, TypeScript’s structural typing created a major issue: engineers unknowingly introduced “fake” SceneComponent objects that had the correct shape but were not real instances, leading to silent failures at runtime. Redux serialization compounded the issue by stripping class methods, causing restored objects to be functionally broken, even though TypeScript saw them as valid. This resulted in unlinked 3D nodes, missing methods, and rendering crashes that were incredibly difficult to debug.
        Unlike Java, which enforces nominal typing and prevents invalid object creation, TypeScript allowed imposter objects to bypass type checks, leading to data corruption and architectural instability. To fix this, we enforced factory-based component creation, branded types to simulate nominal typing, and runtime validation on deserialization. These solutions hardened our system against TypeScript’s weak runtime enforcement, ensuring components maintained their integrity throughout the application lifecycle.