|
isolated experiments and a rank-choice ordering of languages is not a realistic picture of the industry. In robotics, everything is C++. There's plenty of python being used to train networks, but that's not because of performance, safety, or anything, it's mostly born out of the existence of key libraries being in Py. But those libs are just wrappers on C (cuda, essentially). Essentially the whole of every robot is C and C++. Essentially the whole of every airplane is C, C++, and a scattering of memory safe languages in isolated corners. The ATC system, the rail system, most industrial processes are C or PLC, or maybe C / PLC generated by matlab/labview. Automobiles, basically everything with a microcontroller. It's all C. Our scientific computing? fortran. Nodejs? A bunch of C++. What's my point? It's that for any new project or extension of above projects, the existing language is a superior choice (as viewed by managers, business leaders, CTOs, or rushed grad students trying to get quick results, etc), simply because the legacy provides a quicker startup. This is the "reality" - we have had better options essentially forever, and I feel we are effectively stuck with C/C++ forever. It's just that we'll see less and less of it if the new communities are diligent about extending the existing ones. Otherwise, it will never make sense to start clean, not on a mass production level, or at least not this decade or likely next. C++ is a bad choice, and it is the choice. it can be both a prevalent "obvious" choice, and also a bad one. The existence of a better language does not shift reality on its own. You need targeted investment for development of a replacement ecosystem built around that better option. Whether that is Rust, or safe C++, or C+borrow checking, or Dada, or the language of the minute. We as a community cannot keep screaming about how nice a new idea is without building out the ecosystem to make it the obvious idea. |