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.

Minutes as a measure

By Scott Bradner
Network World, 3/16/98

T he Internet is like a packet- oriented device. To the Internet, all is
data and all data is divided into packets. Individual streams of data,
known as sessions, between, for example, a Web server and a
browser are broken up into packets. Each packet contains an IP
source and an IP destination address and is processed individually by
the routers it traverses. In the Internet, the data packets that make up a
single session do not need to take the same path across the 'Net. They
are not guaranteed to get to the destination in the same order as they
were sent. They may be duplicated. They may even be lost and will
have to be retransmitted.

Within sessions, data packets are only sent when they need to be. For
example, if you are using an application that provides for
telephone-like connections over the Internet and you are not talking
(for example, pondering the importance of the last thing that the
person on the other end said), your application is not sending any
data, therefore no packets get sent. Well, if you cogitate too long, a
keep-alive packet may be sent to keep the session alive, but these are
few and far between.

In addition, when you are connected to the Internet, you may have
multiple applications running simultaneously. The Internet protocols
were designed to multiplex many concurrent sessions, not to just run
one at a time. The individual packets have tags to indicate which
session they are part of.

The time factor

There is nothing in the 'Net that lends itself to a time-based
accounting of Internet data transfer. There may be a time-based access
fee, but since a specific connection potentially encompasses many
parallel applications, this cannot be translated into a per-minute usage
fee for a particular Web server.

Billing based on the amount of data transferred may make sense, but
billing based on the amount of time that you spend reading some page
you downloaded does not have any technical or resource usage
justification.

So why is it that most stories about Internet telephony talk about
billing in terms of minutes? Sure it is a familiar concept in the
telephony world because many telephone calls for fax or voice are
billed per minute. For example, there is projected to be over 400
billion minutes of phone usage in the U.S. in 1998. But this does not
translate to use of applications over the Internet. In some cases it
might be nice to have time-based billing - after all, the equivalent of a
page of fax can be transmitted from my machine at home in a small
fraction of a second over my cable-TV based Internet link.

Use-based rather than usage-based pricing would also mean that
sending some packets would cost more than others, a somewhat
strange concept, also easy to defeat by changing the use tag between
consenting end systems. But overall it seems a bit quaint to use
yesterday's billing concepts for tomorrow's technology. The use of
such concepts might even be a way to tell if the pundits understand
the technology.

Disclaimer: Harvard is re-engineering its core systems to avoid the
pejorative use of the description "quaint," but the above are my
thoughts alone.