Your assumption is correct. But you should prepare for the case, that the master goes down, and only the former slave is up. Then your replication should work as well. So you might want have a loadbalancer, which forwards your replication to a running instance.
On the other hand: For what reason do you create a publish cluster? Are you aware of the constraints of a TarPM cluster? Without any strong requirements I wouldn't choose a publish cluster.
Jörg