We use the output paths as keys for determining if index data should be considered "visible". We have a one-to-one mapping between output paths and visible index data. The reason we didn't include this in the target itself is that we expect it to be non-trivial to compute and/or keep up to date. We didn't want to block getting targets on this.
This seems reasonable to me, particularly once we've settled more on exactly what we need from the protocol.