Comment 1 for bug 1341952

Consider a running swift cluster with bind_port unspecified in the object server's configuration file. The same applies to container and account servers, but for the sake of example, let's keep it simple.

Currently, that object server is listening on port 6000; this is implied by the absence of bind_port in the configs. Also, the device specifications in the cluster's object ring all say port 6000; this is explicit in the ring data. There is no default for ports in the ring.

If we changed the default to port 6201 and restarted the object server, it would bind to the new default of 6201. However, the ring still talks about port 6000, so the object server would be unreachable.

While I empathize with the SELinux folks not wanting to allow these ports because of X, I empathize more with the Swift operator whose cluster would be taken down by the proposed change.