I have Transactional Replication set up with a Remote Distributor and a Pull Subscription. So as I understand it, at one end I have the Publisher, and at the other end I have the Distributor, Log Reader Agent and Subscriber. Therefore, all activity is initiated from the Subscriber end. The Log Reader Agent monitors the Publisher, and the Distribution Agent monitors the Distribution database at the Distributor.
Given this architecture, why during the configuration stage does the Publisher need to be able to connect to the Subscriber / Distributor? If the Subscriber / Distributor server can "see" the Publisher, why isn't this enough, as it seems all activity is initiated at the Subscriber / Distributor end?