In Melbourne International Airport about to fly to Wellington; I brought my USB 3G adapter (from three.com.au), which is handy especially in the airport where wifi services are at a premium.

My trip to Wellington is just for the weekend. Going to get the apartment over there ready for some new tenants, and I will pack a bunch of our belongings and send them up to Hong Kong.

My travel this year has been considerable, but somehow I have yet to requalify for AAdvantage Gold. I am about 5k miles away; even though I had so many flights, a large percentage of them were so discounted that they had no mileage accrual.

I had an interesting conversation with a mate recently about Skype supernodes. The question that came up is certainly not new, but it does resurface from time to time: can supernodes in Skype p2p networks create some type of man-in-the-middle attack which may include eavesdropping on transit sessions. This brought me to look at the Skype protocol analysis which was performed by some folks at Colombia University in Sept, 2004. The full paper is here. Now, I am not sure if the protocol has evolved considerably from the time of the paper, or if supernodes are now a commonality in Skype p2p networks. From what I understand, and what the paper describes, if a node is behind NAT or a FW it will not become a supernode. However, there are plenty of network connection methods that will provide a public IP address to a computer. Take some mobile networks for example that do assign from a public pool to nodes. It seems that nodes that have been available for a long period of time, and are therefore deemed as being reliable would be candidates to become a supernode. This is interesting in the mobile Internet model; think about a 1xRTT/2.5G/3G bridge inside a taxi/truck/etc. It could be a mobile supernode that routes calls and messages.

It would be interesting to see a more recent protocol analysis of Skype to see if there are areas that have been further engineered. Since the service now sells commercial calling capabilities, I would imagine that quality of service functions (even if rudimentary) would be developed.