This is an enhancement/usability request for how v2-shim runtime implementations are configured for usage by containerd.

In io.containerd.runtime.v1.linux, the binary name and path to runtime are utilized. As a user, this is straight forward and provides flexibility.

In v2, the shim name is inferred and assumed to be in the hosts' PATH. This is pretty inflexible, and a bit confusing from a usability standpoint. While I appreciate most users don't need to make these configuration changes, I think this could be improved easily.

If not, I would love to better understand the design decisions that lead to the current implementation.