This story appeared on Network World Fusion at

'Net Insider:

Muttering about voice over IP

By Scott Bradner
Network World, 05/07/01            

What is it about voice over IP that creates such passionate dogmatism? Far too much of what passes for debate about the future of voice over IP is dominated by assertions of perfect foreknowledge.

Considering the track records of most prognosticators, I will not add to the conclusions as much as sketch some of the factors that will enter into determining the future - a future I predict will happen in real time.

Debate about voice over IP is dominated by the subject of quality of service (QoS) - at least when anyone who has ever worked for a phone company is part of the discussion. Some have an almost anal fixation on the topic, almost as if they had never heard of cell phones.

The lesson of cell phones - that QoS is not the only factor worth considering - should have been internalized by phone people by now, but that apparently hasn't happened. Too many of them seem to think that no one will use a less-than-perfect phone no matter what other factors might be there. This did not turn out to be the case with cell phones, where the main competing factor was convenience.

In the case of voice over IP, the big factor is cost and will be features. At this point, voice over IP comes at no extra cost - free service might just cure some selectivity.

What factors do affect the quality of voice over IP?

First there is latency. A round trip from me speaking to you and you responding to me needs to be less than about 300 milliseconds. Otherwise, we tend to talk over each other. That latency level is not all that hard to meet in today's networks. The voice coder/ decoders (codecs) can take about 25 milliseconds combined each way, so that leaves about 250 milliseconds for network latency. As I write this, I'm seeing a network round-trip time between East and West coasts of 101 milliseconds (21 hops), so that would leave 150 milliseconds to spare.

Another factor is reliability. Modern codecs can easily deal with 4% to 5% packet loss with no discernable quality degradation.

Because I normally get less than 2% packet loss on the 'Net, I should get quality at least as good as a normal phone and better than a cell phone.

So what's wrong with this picture? I'm working late Saturday night - the results might be rather different at 2 p.m. on a weekday.

How about prioritizing the voice data on the 'Net? That might help quite a bit, but why should someone other than my ISP honor the prioritization - because there is no way for me to pay them to do so?

Then there are the regulators. Voice over IP is against the law in more than 40 countries because it takes revenue away from the telephone companies (and taxes away from the governments).

Not a clear picture, but "free" is a factor to remember.

Disclaimer: Harvard does not generally comment on free things and has not in this case.

All contents copyright 1995-2002 Network World, Inc.