Internet neutrality: What is fair to the end user?

Today, we respond to another reader’s question about our recent coverage of the Internet neutrality issue. The reader writes: “While an undifferentiated best-effort service is fine for accessing the Web and sending e-mail … real-time services such as VoIP require more predictable behavior if they are to work well. So the choice is between (a) unreliable VoIP (fine for a free supplementary service such as Skype, but not a satisfactory substitute for the PSTN); (b) differentiated services, the current bete noir; or (c) a network so fast that no significant statistical multiplexing occurs.”

Our reader continues by suggesting that although ‘Net neutrality proponents might like option c, it really isn’t an economically viable choice. Rather, he suggests, that the only way to offer “reliable services such as VoIP over a best-effort network is to severely limit the amount of traffic allowed into it, which brings up all the same concerns about fair access to resources as differentiated service does.”

As we and others have noted, the need for prioritization of real time services like VoIP and video needs to and is happening on the IP backbone and at the edge. And as our reader points out “once the need for differentiated services is accepted, then some means is required to stop all the traffic getting loaded into the ‘top priority’ class.”

He concludes that it would be better to make sure “that the mechanisms deployed to provide differentiated services don’t impose any penalty when they’re not needed [rather] than to try to turn the clock back to the time when best effort was good enough.”

We agree that real time traffic must be prioritized and that it makes sense for a user to pay more to transmit a statistically-guaranteed voice bit than for a best effort data bit. But we still don’t think that the called party (or the content provider) should pay extra to the access provider for us to use their content or application. However, we do think that if users want to guarantee their own QoS, then it may be better for the user (or the calling party) to pay for better or faster access.