There are plenty examples of why HTTP is great for communicating with clients, but I think Miles is wondering more about communication between one couch-node to another.
The case for HTTP is less clear in that case, but the lack of clarity might only be temporary. Besides the direct benefits of shuttleing bits over plain old HTTP within and between data centers, I can imagine one or two scenarios where replication could be used for communicating with something other than another couch-node (however ill advised that notion might actually be). -R
