At first the network neutrality issue for broadband Internet service was all about the desire of carriers to realize value from content that customers wanted to download.

Spokespeople for carriers such as AT&T talked openly about charging content providers such as Google extra to get their content to end users -- in effect charging a third time for the content, since the end user and the content provider were already paying for their Internet service. Amazingly enough, the carriers have proved smart enough to realize that this approach was just getting them into deep doo-doo and have started to change their approach.

Now they are focusing on dealing with bandwidth hogs and implying that all will be right with their world if these bandwidth abusers would just go away. Since the solutions being proposed generally will not solve any actual problems that the carriers have, one has to wonder if they have some other motive.

We have to take as a given that no network can be built in such a way that all end users can use their connections to their full bandwidth simultaneously. It is not economically, nor generally technically feasible. All networks are built with some level of oversubscription, with the assumption that not everyone will try to use their full bandwidth at the same time.

The level of oversubscription varies greatly between carriers and varies at different points within the carrier’s network. For example, one point of oversubscription for cable-based carriers is the local cable loop, and a common point of oversubscription for DSL-based carriers is the uplink from the DSL multiplexer in a POP and the rest of the carrier network.

These points of oversubscription become congested if too many customers try to transfer too much data at the same time. In some cases it is easy to increase the bandwidth at these points to reduce the level of oversubscription, but not always.

Telephone networks solve this type of problem by not letting people make new phone calls if the network is congested -- you get a fast busy signal. There is no equivalent to a fast-busy for the Internet -- that is, there is no distributed admissions control -- so there is no way for the end system to know if some points in the network are already congested and to not try a big data transfer. The Internet mostly relies on congestion-aware protocols, such as TCP, which slow down in the face of congestion. But not all Internet protocols are congestion-aware. For example, music or video-streaming protocols tend to ignore congestion.