I'm reading this whitepaper and trying to wrap my head around why it would require O(N^2) entries for MAC address learning in OpenFlow with single table. See paragraph 2.2.1. In my naive mind, I think for every MAC address you would basically have an entry that says:
Source: *, Dst: <some MAC address>, out_port: <port that MAC was seen/learned on>
so you would end up with O(N). What are the other required entries?