DPI and Net Neutrality's Overseas Weak Spot
Ian Lamont writes "An unnamed source at an American ISP says staff there briefly considered using Deep Packet Inspection to comply with an order from Argentina's Department of Justice to block access to a local gambling site. The ISP ended up not going that route, owing to the cost, but some engineers at the company worry that DPI will eventually be implemented on the ISP's overseas network, thereby positioning it for an easier US rollout should Net Neutrality lose out in Washington. Besides being used for traffic-shaping, DPI can also monitor the traffic of ISP subscribers to supply targeted advertising."
And say "No".
Even if it hurts in the short run. The loss of consumer bargaining power in these instances, where the contracts possibly allow for this, is the fault of the general consumer to begin with.
IMHO Deep Packet Inspection will be rolled out to identify the protocols in use on connections, to support assigning the correct QoS to different protocols.
For instance: File transfers accelerate until they consume (and equally divide) all bandwidth at the most congested link in their path, but just slow down if they're artificially limited below that level. Meanwhile Streams are band limited but must go to the front of the line to meet their jitter and delivery reliability requirements, though delayed stream packets are useless and should be dropped to avoid also delaying their successors.
Unfortunately the tagging of the packet itself can't be trusted because there is an incentive to achieve improved service by cheating, requesting better service than necessary. (And a Microsoft IP stack, widely deployed, made just this "improvement".)
My take: The right solution is to write a contract for various rates of "premium" packets, then accept the labeling but demote the QoS on packets above the running limit. Then the incentive is on the user to obtain software that doesn't cheat, and the ISP doesn't need to deep inspect.
Unfortunately, the ISPs and equipment vendors seem to be going with the DPI identification approach. And that means deploying DPI, which can then be misused by the ISPs to do the bad kind of non-neutrality.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
They throttle https? How have online banks and retailers reacted?
I'd hand out a complimentary tinfoil hat if I had one.
IPv6 is on the radar and requested as a must-have, but normally only on a roadmap level ("Will your product support this some time in the future?"). In some parts of the world (there's more to it than the US), any device incapable of IPv6 won't get onto the network in the first place.
If you stop to think about the practical implications for a while, it's very unlikely that encryption will be that much more widespread than it is today (it's a processing power issue as well, not just one of protocol ease of implementation) while the whole NAT issue will be zapped. This means that DPI gear all of a sudden can pick out a whole lot more, since traffic that'd normally be aggregated by a NAT - won't be. Insta-higher-resolution.
There's no conspiracy here. Really.