There is clearly a possibility of causing a denial of services using this kind of 'attack'.
In theory on would make allowed address pairs a resource with a quota, but that's not easy as they're an attribute rather than a resource.
So I would rather take the approach in the suggested patch of putting a configurable hard limit on the number of address pairs allowed on a port.
It should also possible to have both a maximum number of pairs per port and a global per-tenant maximum number of additional address pairs.
There is clearly a possibility of causing a denial of services using this kind of 'attack'.
In theory on would make allowed address pairs a resource with a quota, but that's not easy as they're an attribute rather than a resource.
So I would rather take the approach in the suggested patch of putting a configurable hard limit on the number of address pairs allowed on a port.
It should also possible to have both a maximum number of pairs per port and a global per-tenant maximum number of additional address pairs.