Sample Performance Results
	We show application level performance by observing voice stream arrival delay. At the mobile node, we kept a log file to register the arrival time of each 144-byte block (talk burst) in the voice stream. The figure plots the arrival times of the first 20,000 blocks at the MN from two remote nodes: (a) Texas and (b) Virginia. As we can observe, the IPMN dramatically outperformed Mobile-IP on two levels: Firstly, in general, most blocks were delivered faster with IPMN due to shorter triangulation-free path that they had to travel to reach the mobile node as well as to smaller overhead. Secondly, IPMN plots were much smoother than MIP since the latter suffered from longer disruption of TCP service due to longer handoff delays. After each handoff event, we see the impact of TCP's slow start behavior on the plot. These step jumps and the impact of TCP dynamics created jitter on the voice stream.
	One of the key features of our interactive scheme is short handoff latency. After running the experiment several times by placing the correspondent node at (1) locally, (s) in Virginia, and (3) in Texas, we observed a big difference–up to two orders of magnitude—in handoff latency between IPMN and classic Mobile-IP. The table shows the handoff latency of the first five handoffs on the three nodes for both Mobile IP and IPMN runs. IPMN managed to perform handoff in 110 to 200 milliseconds on average while Mobile-IP needed between 14 to 44 seconds. This substantial reduction in handoff latency highlights the advantage of event-based protocols like IPMN over timer-based protocols like Mobile-IP. The former allows protocols in different layers to interact and pass events and new state information–like the new IP number in our case—to upper layers instantly. This enabled peer protocols to respond immediately, therefore cutting down unnecessary overhead time. Timer-based protocols on the other hand usually use a timer-based periodic probing mechanism to discover state changes. For example, in this particular implementation of Mobile-IP that we have tested, the foreign agent sends beacon signals (agent advertisements) to discover mobile node movement every 60 seconds! A best case scenario will happen if LL handoff was performed right before the arrival of a beacon signal. Therefore, this process will take half of that time on average–i.e. 30 seconds. Adding to that communication and address registration overhead we can easily reach the 40 seconds average especially on the two remote nodes.