| My application is a bunch of docker containers networked together using docker compose. Now I want to ship this entire stack as an easily runnable application to my users, who are either on Linux or Windows (and occasionally Mac). Ideally the user doesn't even know that docker is involved, and it just looks like one app from most perspectives. I want it to be easy to install and maintain. To further complicate things, many of the containers are actually dependent on nvidia-docker and GPU access for CUDA. This seems to just now be supported on Windows, albeit only Windows 11. In general, this greatly complicates my ability to ship anything easy-to-install. Is there anything I can do to make this process easy for users (and myself)? NVIDIA and docker have both taken over huge swaths of our computing infrastructure, but they are in some ways a huge regression for easy software delivery, which is somewhat ironic in the case of Docker. | 
reply