The following text is copyright 1998 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.

It's a curve, not a point

By Scott Bradner

Network World, 6/29/98

The Internet is not good enough, or so many people would have us

The quality is not good enough for IP telephony or IP-based virtual
private networks (VPN). The security is not good enough for remote
access or IP-based VPNs. And the reliability is not good enough for
sending faxes or conducting electronic data interchange over IP. All in
all, if you listen to some pundits, you'd think the Internet isn't of
much use for anything real.

There is a theme to these pronouncements. They assume that there is
actually some single point of service acceptability. Because that has
repeatedly been shown not to be the case in the real world, it is
reasonable to wonder how much real world understanding these
pundits have.

This is not a new misunderstanding. A number of years ago, one of
the telephone company standards groups did an exhaustive survey to
determine the minimum quality needed in the telephone system. This
involved a lot of experiments and survey work and is frequently cited
as a landmark study of its type (though I've not been able to track
down a copy). My problem with the work focuses on the conclusions
drawn from the effort, which determined a point of acceptable quality.
This conclusion shows a misunderstanding of the marketplace's

For many years, most traditional telephone companies refused to
invest in the infrastructure to support cellular phones because the
voice quality did not meet the acceptable minimum. The assumption
was that no one would want to use the service.

It was only after the frequency spectrum opened up and new
providers were able to show that customers wanted the service that the
mainstream phone companies made their move into the market.

Those people who say the Internet is not good enough for one
application or another are missing the same detail that the phone
companies did. There is not a single point of acceptability, but rather a
curve of quality vs. other factors. These could include: convenience,
as was the case with cell phones; cost; service coverage; ease of use;
and many more.

CNN ran a story on IP telephony a week ago in which a user of the service was interviewed. This is a service that involves dialing the number of a local gateway from your regular phone, then having the call routed over the Internet to a gateway near the call destination, where the call is put back on the local phone network. The user said that there were a few problems with the quality of the service, but he seemed quite happy with the cost-quality tradeoff.

You should look closely at any pronouncement that a particular service cannot be successfully run over the Internet and see if there is a missing curve in the assumptions. Remember the cell phone.

Disclaimer: Since Harvard does not make pronouncements and this is one, it must be my own.