What is the roadmap for copy free ROS messages with GPU data on NVIDIA Jetson

When using ROS messages on a single device, the data, including GPU data, should be shared, no data copy should be required. While deepstream provides a copy free GPU data pipeline, my understanding is that ROS messages do not yet. Is NVIDIA, as part of its initiative to integrate ROS with Isaac, going to extend ROS messages to pass on GPU data without copy? What is the timeline?