Saul Hansell of the New York Times has posted an interesting discussion with Kurt Dobbins, the CTO of Arbor Networks. Arbor is one of several hardware vendors working on deep packet inspection (DPI) hardware for a number of uses, though Dobbins prefers to downplay the technology's more seedy applications -- instead focusing on how DPI can be used to charge broadband users more money. He's particularly interested in pricing models in the UK:Thankfully, Hansell quotes broadband-industry mainstay Dave Burstein, who again steps up to remind people that ISPs already make a killing on flat-rate pricing plans, and that many of these fractured, overage-oriented plans are absolute consumer ripoffs. Most usage-oriented plans have less to do with network capacity, and more to do with placating hungry investors, for whom ridiculous profits are never enough if there isn't staggering quarter-over-quarter improvements. A desire to protect TV revenues from Internet video plays a role as well.

As for Deep packet inspection, it has many legitimate uses on an ISP network, but has gained notoriety in recent years for its use in delivering behavioral ads, injecting ISP javascript banners into websites, and identifying (and ultimately filtering or throttling) P2P traffic. Evolution in DPI technology has companies not only buying and selling your online habits, but pushing the technology as a new Internet policing solution, potentially placing the onus on ISPs to compare every packet a user sends to a master database of what's legal or illegal.

While Arbor may not want to talk about it, DPI's creepier implications led consumer advocacy group Free Press to last week issue a report (pdf) on DPI, warning that while the technology itself isn't necessarily bad -- the way in which it could be mis-used by carriers may be. Last fall University of Colorado law professor Paul Ohm made a similar argument, saying the modern ISP, thanks to new technological developments like deep packet inspection, poses the greatest privacy threat to American consumers.