Show HN: React WebGL component: 45 3D formats viewer.Should I industrialize it?

Babylon.js or Three.js are pure javascript libraries. Unlike Unity, they are light and do not require an IDE. They are free and open source. By the way, I'm thinking of compiling the loader for Babylon or Three. I chose Unity here because I master more and because the integration of C++ (for the libs I bought to decode some formats) is simpler. I also chose Unity because I would like to use my loader to create an augmented/virtual reality 3D model viewer application for Hololens, Open VR, w10 Headsets, ... And Unity can compile the same code for targets that have nothing to do with each other.

Not the OP, but having used Babylon.JS it works fine for simple things but the codebase is a disaster, particularly anything physics-related. It's clear the architects and devs had never worked on a serious professional engine before starting working on it because the codebase is riddled with rookie game dev mistakes. The biggest is a complete failure to understand the importance of dimensions and units in the camera and character movement code and physics systems. Distances are added to velocities, directions are used as accelerations, time is used as dimensionless factor, if you've ever done serious Physics coding what you find in the codebase is terrifying. There is no way to make the system frame-rate independent as a result because what you're really doing when you garble units like this is making assumptions that the frame rate is constant so time and inverse-time factors can be ignored (hello, 1990's calling! We want our bugs back!). Once this kind of stuff gets embedded in a codebase it's very hard to take out on a piece by piece basis because all the math in your architecture ends up being a hack that happens to collectively mostly work. As you start trying to remove it, the hack stops being mostly self-supporting and starts breaking everywhere until you complete the fix, but there is no way to know where all the flaws are without an exhaustive single-pass code review of the entire codebase by a serious skilled and experienced dev or devs, which is unlikely to happen (particularly when the core mantra of project leadership is "100% backwards compatibility always!" which is a great concept until you find something you really screwed up originally and now you're stuck with it for eternity).

Huge kudos to the team for getting bought by Microsoft. Yes I'd use it again on another very small "show a model but don't need any serious gameplay" project but I'd be very scared of anything where I was intending to invest more than a tiny amount of coding time in the project because the architectural and implementation failings will balloon your coding time dangerously.

Assimp, TriLib and some Unity native features