Does redf replace nodl, or could it merge with it?

[sloretz]

A while back Canonical started Nodl, but never really finished it. It seems similar in purpose to redf.

• https://github.com/ubuntu-robotics/design/blob/node_idl/articles/ros2_nodl.md
• https://github.com/ubuntu-robotics/nodl
• https://github.com/osrf/nodl_to_policy

Does redf replace nodl, and if not could redf and nodl be combined?

[koonpeng]

This started as a result of a discussion with @Yadunund and @codebot. I wasn't aware of nodl but it looks really interesting. I think there are some slight differences currently but it would be nice to have a universal descriptor format that handles both use cases.

[clalancette]

I do think it would be a good idea to see if we could adapt NoDL for whatever use case redf is being used for. That way we won't have 2 new descriptor formats that are doing similar things.

[Yadunund]

One of the goals of redf is to auto-generate a library that downstream packages/nodes can include to get the names, qos, etc for an interface. In a way, it is used to configures the ROS 2 interfaces on a system level. Nodes in the system will include the same package to the get properties of a given endpoint. The motivation of this is to make development of ROS 2 nodes easier for new system integrators by saving them time to lookup what the right endpoint names and qos should be.

nodl seems to be at the node level? ie each node has an accompanying descriptor of its interfaces which can be used to statically analyse whether a second node is compatible to communicate with the first?

@clalancette and @sloretz do you think there is value in pursuing redf as a tool that can auto-generate this utility "interface library"? We can certainly default to using the descriptor format specified by nodl and not redefine that here.