Speaking out on voice-overs

There has been a lot of buzz of late about voice-overs. Voice over DSL and voice over cable are all the rage in the trade press. But much of the buzz misses what may be the real story.

Standards are being defined on how to run voice over these technologies. The model assumes the service provider has a DSL circuit or cable connection from a point of presence (POP) to a customer location. At the customer location there is a modem-like piece of customer premises equipment (CPE) attached to the line. The CPE has multiple connectors, one or more for each service that can be offered.

For example, a cable modem might have connectors for cable TV, high-speed Internet service and one or more phone ports. At the POP, the other end of the line is attached to some electronics that split the customer traffic into separate streams, one for each of the connectors on the CPE.

In this model, if the customer wants voice service he plugs a phone into a phone port on the CPE and the service provider configures the POP equipment to logically connect that phone port to a telephone service provider. The customer appears to have a direct connection to the phone service provider and gets the same type of phone service and bills that any other customer of that phone company would get, along with some potential savings resulting from the use of a shared-access line. This model is quite attractive to service providers because they can bill for each service.

But this is not the only model. An alternative is where the CPE is simpler and only provides cable TV and Internet service. If the customer wants to make phone calls, he does so from an Internet-enabled phone or a PC. In this case, the voice traffic is not separated out; it is sent along with all the other IP packets on the Internet link. The phone call can go over the Internet directly to someone with another Internet-enabled phone or to a gateway to the phone network run by a third-party.

This is not attractive to service providers because there is no reliable way for them to figure out which is voice traffic and charge separately for that. But it will be attractive to customers for exactly the same reason.

The service providers may claim that without separating the traffic, they cannot guarantee the right quality of service for the application. I would claim that you should not be buying Internet service from a provider whose basic service is crappy enough that this makes a difference.

In any case, this alternative model has been generally ignored in all the buzz, but may easily be the dominant model in a few years.

Bradner is a consultant with Harvard University’s University Information Systems. He can be reached at