“œWe’re becoming an internet service provider!” observed a mobile operator during a recent visit to the Cisco briefing center. The comment stemmed from a discussion about how the mobile internet keeps growing more pervasive in their technical designs, business decisions, and even cultural approach. And as with any sweeping change, the epiphany comes with a mixture of excitement and trepidation.Wireless flirted with the internet since its beginning. In 1970, as the ARPANET (internet ancestor) first connected across the U.S., ALOHA packet radio emerged in Hawaii. But the ALOHA access techniques soon left the airwaves to form the foundation of Ethernet, and for the next two decades the internet grew as a fixed network. (For a fascinating review, check out Leonard Kleinrock’s recent IEEE Wireless Communications article,”History of the Internet and Its Flexible Future“.) Wireless and the internet reconnected in the 1990s, as WiFi and 3G started delivering IP packets at useful speeds. Faster radios, lower charges, and attractive devices have thrust the mobile internet onto center stage.Creating the mobile internet isn’t as simple as sending IP packets over radio links. The data traffic flows differently than telephone calls, and differently than the assumptions built into the initial standards, which call for an elaborate hierarchy. (To the tune of”Dem Bones”: The ME connected to the BTS, the BTS connected to the RNC, the RNC connected to the SGSN, the SGSN connected to the GGSN-oh, hear the word of 3GPP!) This hierarchy best supports a centripetal flow of traffic, assuming users would mostly browse”on deck” content in the mobile operator’s data center. But in fact, much traffic is destined outside the operator’s network, particularly encrypted VPN traffic from users’ laptop cards.These network loads are prompting upgrades to the mobile infrastructure, like that recently announced by 3 Scandinavia. The centrifugal traffic patterns are also prompting changes in architectural direction, distributing mobility functions towards the edge, often described as a”flat” IP network. In the extreme, this trend would move all mobility functions to the cell site, but that would makes sense only with limited mobility (think laptops in cafes) and plenty of bandwidth (fiber) to the cell site. Otherwise, the churn of connection hand-offs would overwhelm the RAN transmission links. Still, the trend towards”flatter” IP networks moving mobility functions towards the edge seems likely to continue, driven by reducing the expense of unnecessary traffic backhaul.Traditional internet service providers would recognize this approach as”hot potato” routing, where packets are forwarded to other operators’ networks as soon as possible. In this mindset, traffic is an expense to be deflected, not a treasure to be hoarded, a mindset more usually reflected in the construction of”walled gardens.”Another example of changing technical and business mindsets comes from WiFi on dual-mode devices like the Apple iPhone. In early planning, some mobile operators sought to disable the WiFi connection, to avoid”losing control” of the customer. But consider someone web browsing with an iPhone in their home. Why wouldn’t you want that traffic to travel over their WiFi and broadband connection (at their expense), rather than burdening your 3G network resources? A flat-rate data plan yields the same revenue either way. No wonder AT&T cut a deal with Starbucks: one more way to reduce the cost of connecting iPhones and other devices.Economics if not predilection pushes mobile operators to consider flatter IP networks, incorporate WiFi and other access technologies, and revisit their architectural assumptions as they plan for the mobile internet. Expect more epiphanies to come as mobile operators also become (at least in part) internet service providers.