The following text is copyright 1997 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.

A distorted economy

In his editorial a couple of weeks ago John Gallant talked about the imminent major change in the telcom world based on the development of the "true integration of packetized voice and data across frame relay and ATM, as well as intranets and the Internet." I wonder just why this change is so inevitable.

While it's a very attractive philosophy to integrate all your communications support over one infrastructure, it does not make much sense using current generation technologies. It may not even make technical sense in the long run. Yet the path to global service integration is painted as a golden one.

Voice and data make very different demands on a connecting network. The very tight delay variation tolerances required in the interactive voice world are in conflict with the demands on a packet-based data service in which delay variation is far less important than delivery reliability. Voice connections tend to be long duration, multiple minutes being a short call where the average data connection on the Internet is less than a dozen packets and is finished in a few seconds. A network built to support voice traffic well will be different and more expensive than a network built to support data. Carrying data over a combined network would be more expensive than carrying the same data over a data-only network.

If the amount of data to be carried is relatively low the additional cost could easily be less than the savings resulting from having a single physical infrastructure. But in a few years there will be more data bits transferred over the telecommunications systems than voice bits. A higher percentage of data traffic will reduce the comparative benefits of integration, eventually to a point where integration costs more than it saves.

John talks about integrating "packetized voice and data traffic on more cost-effective carrier/Internet service provider services." Putting aside for a moment the issue that we do not currently have deployable technologies to support this type of integration over wide area networks (and can only support it on LANs by installing ever faster networks such as fast and gigabit Ethernets) how can it actually be true that the packet services are more "cost-effective"?

It can be (and is true) not because of any better efficiencies of packets as compared to circuits but because the price of voice service has little to do with the cost of providing the service. The price has been manipulated by regulations promulgated to provide universal service, life-line service, and a myriad of other "good" things and is inflated by federal, state and local taxes and FCC mandated line charges. In addition the phone companies have been operating in a regulated environment where an additional cost to provide the service translates into additional profit due to the guaranteed profit margins.

I predict (sadly) that if this integration process reaches any significant scale those who tapped the phone revenue to do good deeds will find a way to maintain their revenue stream and thus negate much of the savings from any integration.

disclaimer: Harvard is an aggressivally decentralized organization and, so far, does not relate well to integration concepts so the opinions are my own.