Tuesday, July 24, 2007

Net Neutrality Explained!

TCS Daily has yet another silly post on net neutrality. It's depressing how many people opine on this without actually describing the real problem. There are two different ways that an ISP can provide unequal treatment to the packets flowing through their network:

1) They can discriminate on all the packets coming from or going to a particular address. For example, an ISP might give preferential treatment to all of the packets coming from Google, but give only best-effort service to packets coming from Time Warner. This would obviously give Google a competitive advantage.

2) They can discriminate based on the type of packet. For example, they might charge differently for streaming video or VoIP packets, which require high-bandwidth, realtime delivery, than they do for email or IM packets, which can delivered best-effort.

Now I have yet to find anybody who thinks that discriminating based on customer's identity is a very good idea. Nobody's doing it today (as the authors point out), but it would certainly be possible and would probably give big companies a way to ice little companies out of their markets. Do we need regulations to ensure that this doesn't happen? It should at least be debated.

On the other hand, allowing ISPs to charge differently for different types of packets--what's called "class of service"--is absolutely essential. Most ISPs are still blessed with abundant bandwidth and excess router capacity, but that will change soon as high-bandwidth, time-dependent applications become more popular. ISPs simply cannot deliver realtime video and audio reliably without reducing the load on some of their routers. This means they have to buy more routers, switches, and network links, which implies that they have higher capital and operational expenses to handle these types of traffic.

Regulation that prevents ISPs from recovering their costs on these differentiated services will simply prevent those services from working reliably. It'll be a disaster.

No comments: