In a conversation with @prlawrence and my thread with @AastaLLL, they said that JP6 would be more flexible and simpler to do custom integrations with distros.
For the first time, JetPack 6 brings the flexibility to run any Linux Kernel and offers wider options of Linux based distros from Jetson ecosystem partners
My interest (from that thread) is primarily around the ability to run containers on a different distro, not actually JP6.
Is the process of getting the dependencies to make this work, described in the other thread, different for JP6 than JP5? That process was:
The list of files necessary to be mounted into the container
The process for creating the CDI so that containerd can consume it
The installation of the dependent files (mostly .so libs)
From the point of view of the NVIDIA Container Toolkit, there are no differences between various JetPack versions. Both platforms provide CSV files (l4t.csv, devices.csv, drivers.csv) that define the entities that are required for containerization. The NVIDIA Container Toolkit consumes these and generates a CDI specification (and if used in the context of the nvidia-container-runtime directly applies this generated specification to an incoming OCI runtime specification).
As pointed out in your other thread, the exercise for supporting a new platform / distribution is then broken down into:
Collecting the required entities (e.g. .so libs)
Constructing (or reusing) the relevant .csv files
Generating the CDI specification from the specified files.