Dr. Denis

@fmauch and @gavanderhoorn this is also a very plausible solution. On the other hand, as for the controller stopper, this could be one of the additional functionalities of controller manager....

> We were discussing making the controller_manager a lifecycle node. In this case, spawner could simply depend on controller_manager in the launch file so it only tries to spawn controllers...

This can/should be done using hardware lifecycle so that you change state from active to inactive. This should be done by returning 'ERROR' in read or write and handle it...

> **Edit:** What I missed earlier: If we simply return `ERROR` from write, we're effectively back to #262 again. Exactly, as I mentioned, we will need to add some more...

Yesterday, the last missing piece to do recovery manually landed in the ros2_control's main branch. So the approach to handle this is the following: 1. Hardware returns error on read...

@livanov93 Is this sill used?

> > Does this mean that ros2_control does not support cartesian commands _at all_? As in it is not possible to define something like a `System` hardware component that offers...

> Was something like this that you had in mind? Yes, exactly! You should add other interfaces for Cartesian commands, and in `prepare_` and `perform_command_mode_switch` you can detect mode you...

The package is not release so you will not find it in ros-index for now.

I turned this off at one point because it didn't work...