Policy Definition should use port names rather than port id

[gmcnaught]

Today in the configuration we see something like this below:

rule 6182: FlowEntry[dpid=[00:00:00:a0:a5:7a:d7:34],ruleMatch=[OFMatch[in_port=59590,dl_vlan=4092]],actionsList=[Slice:nddi=7],id=[7200],priority=[10],]
rule 6183: FlowEntry[dpid=[00:00:00:a0:a5:7a:d7:34],ruleMatch=[OFMatch[in_port=59590,dl_vlan=4093]],actionsList=[Slice:nddi=7],id=[7201],priority=[10],]
rule 6184: FlowEntry[dpid=[00:00:00:a0:a5:7a:d7:34],ruleMatch=[OFMatch[in_port=59590,dl_vlan=4094]],actionsList=[Slice:nddi=7],id=[7202],priority=[10],]

There are 2 problems with this approach for us: First, we have seen cases where the id assigned to a port changes over time because these are ephemeral values, If in production the port id changes this causes policy failure. Second, for engineers having to debug policy in production port ids and dpids are too low level, these should be defined and displayed using interface names eth1/1 etc. To the best of our knowledge the spec does not address how stable the port_numbers should be.