The Kubernetes client, CD driver and owning resource are pretty much propagated everywhere, which adds a lot of bloat to function signatures, structs that have to cache these values etc. What would be a lot cleaner is to pass these about in the context which is by necessity propagated everywhere. The added side benefit is we can override the values in different scopes, for example, if we use a remote instance of Argo we can change the client to the remote cluster, and the driver to the remote Argo instance.
The Kubernetes client, CD driver and owning resource are pretty much propagated everywhere, which adds a lot of bloat to function signatures, structs that have to cache these values etc. What would be a lot cleaner is to pass these about in the context which is by necessity propagated everywhere. The added side benefit is we can override the values in different scopes, for example, if we use a remote instance of Argo we can change the client to the remote cluster, and the driver to the remote Argo instance.