Nolle's column: The truth about voice over IP

Carriers make over $200 billion annually worldwide on voice communications. Despite claims that data traffic is exceeding voice traffic, voice revenue is what's driving the carrier market. Naturally, the data vendors want a piece, so there's a popular trend to link voice with the Internet.

But do users believe in voice over IP? Will it work technically? Is the regulatory environment favourable? What will happen to the Internet if voice over IP succeeds?

In a survey of 267 companies of all sizes I completed lasst year, respondents didn't see the need for voice over IP to augment or replace basic local, long-distance or internal telephony. More than 95 per cent of the respondents had no plans to replace general domestic voice calling with voice over IP. A mere 9 per cent were interested in voice over IP for purely intracompany tie-line voice. Even for international calling applications, where voice costs are high, only 19 per cent of respondents said they would consider voice over IP. Straw polls I conducted at seminars on the East and West Coasts within the past month confirmed this view.

The reason companies most often give for their lack of interest in voice over IP is the savings vs degradation issue. Most large companies already have voice rates below 7 cents a minute and believe competitive pressure will bring prices down further. Nearly all believe that the benefits of voice over IP will be insufficient to justify the speech quality problems.

Research on voice over the Internet has shown callers can expect round-trip delays of 500msec or more - double the delay of satellite. This delay is intrinsic to the Internet's structure, so the voice-over-IP players can't control it. In my survey, delay-induced problems with speech topped the list of user complaints about voice over IP by a large margin; 77 per cent reported delay unacceptable, while only 17 per cent reported problems with packet loss or compression-induced distortion.

Even if voice over IP worked, would voice over the Internet be effective? The truth is that most voice-over-IP applications depend on the concept of unlimited-usage pricing of IP service. Traditional dial-up voice services are charged by the minute and sometimes are also based on distance. Internet voice traffic usually falls under the monthly fixed rate, so ISPs have to handle more traffic with IP voice without reaping more revenue. Enough IP voice and the ISPs would lose money.

We might want to consider support costs when we assess how earnestly carriers would promote voice over IP, based on recent voice-over-frame relay experiences. Many carriers have quietly begun to ignore requests for proposal for voice over frame relay. The reason, according to one carrier, is customer support costs are eight times that for frame relay data because of customer complaints about voice quality. Support costs for voice over IP probably would be even higher.

Even if the ISPs could make money on voice over IP, regulatory issues loom. The US Federal Communications Commission has indicated it will, in some cases, impose access charges on ISPs to fund the universal service pool. Because that pool is targeted at subsidizing voice service, it's clear ISP entry into the voice market is one of the cases that would likely induce FCC intervention. In other countries, where voice services are the monopoly of a national carrier, large-scale voice traffic migration to the Internet would surely result in an order to cease transport of voice traffic.

That might be a good thing for everyone on the Internet, because there's reason to believe that the success of voice over IP could literally shut the Internet down. In the late 1980s, the Internet suffered a series of congestion collapses that arose from the way TCP flow control and error recovery handled network faults. The introduction of the "slow start" feature was designed to correct this problem; it is now included in all TCP implementations. But voice over IP can't tolerate the delay induced by slow start, so it doesn't use TCP. If voice-over-IP traffic on the Internet grew substantially, a large part of Internet traffic would be operating without slow-start flow regulation, and the Internet could be driven to the same kind of collapses we had in the 1980s.

So what's the truth about voice over IP? First, users are almost universally doubtful about its potential in any form. Second, there are a host of technical, business and regulatory reasons why it probably can't be broadly useful on the Internet.

What about private networks or new carriers such as Qwest Communications? Could voice over IP be more efficient in that context, where presumably delay and quality of service could be controlled? Perhaps. Compressed voice requires less bandwidth than standard voice - as much as 85 per cent less. Couldn't that cut costs by 85 per cent?

Not according to financial analysts, who point out that the cost of bandwidth is only about 10 per cent of total long-distance telephony cost. The rest is switching, customer service and support, and other expenses.

Voice over IP is useful for Internet call centres, collaboration between Microsoft NetMeeting users and cell-phone partners trapped on expressways and other valuable but niche applications. In the past, when we've over-hyped a technology, we've succeeded in killing even its legitimate uses. Let's be honest with ourselves and give voice over IP a chance where it's real.

Thomas Nolle is president of CIMI Corp., a technology assessment firm in Voorhees, New Jersey. He can be reached at +1 (609) 753-0004 or