InfoMagic Standards 1993 July/Disc.iso/inet/isoc/isocnews/issue1_4/070.10a

Internet Message Format | 1993-02-08 | 5.2 KB

Subject: n-1-4-070.10

The Internet Society President Replies


[ed. In a recent issue of Info World, publisher Bob Metcalf wrote
an editorial based on a luncheon meeting with Internet Society
President Vint Cerf.  Bob, who developed Ethernet and founded the
3COM Corporation, is also an old friend who enjoys making probing
remarks - in this case about the Internet world.  The following
reply by ISOC's President is a useful primer on the multiprotocol
character of the Internet, as well as the much broader base of
internetworking technology in general.]


The Internet is a global phenomenon and, in its present form, it
supports a multiprotocol environment which includes TCP/IP but also
CLNP from the OSI suite and a variety of proprietary protocols.
The protocols of the TCP/IP suite are by far the most uniformly
distributed in the system and the TCP and IP layers are often used
to support "tunneling" of other protocols across the Internet.
In addition to the million hosts on the global Internet, some 30,000
other network numbers have been assigned by the Network Information
Center. These nets are not on the global Internet but form a collection
of private internets.

To characterize the Internet Society as a cult is silly. There are
millions of people who make use of the technology either on the global
Internet or in private internets. Their interests range from the
evolution of the protocols, which are changing and adapting to new
demands, such as packet audio and video multicasting and multimedia
email, to applications, to economics and sociology, English literature,
physics, chemistry, biology, and astronomy. It's simply a cross-section
of people in research, education, and, increasingly, the business
community. Interest in the Internet and its technology is based on
real utility and excitement about a system which accommodates innovation.

With respect to OSI, and in particular X.400 and X.500, you might not
be aware that much of the X.400 email on the Internet runs above the
TCP/IP protocols (a thin TP0 emulation layer - see RFC1006 - is used
to accomplish this). The Corporation for Open Systems, a founding
organizational member of the Internet Society, has linked the OSINet
to the Internet to allow both systems to be used to support tests
of OSI protocols in the Internet. CLNP is supported on the NSFNET
backbone and in a number of other constituent networks.

In the multiprotocol environment, it seems to me perfectly reasonable
to continue the exploration and evolution of the TCP/IP protocols.
They have been adapted to operate at gigabit speeds, some excellent
work on multicasting and flow management in the routers has opened
the door for audio and video transmission, multi-user games and work
group applications are emerging. This is a vibrant development
environment.

With respect to privatization of the Internet, it is already happening.
A number of for-profit Internet service providers have emerged such
as Performance Systems International, ANS Co+RE, UUNET Technologies,
CERFNET [I am not connected with this General Atomics subsidiary],
Sprint International's Sprintlink, Infonet's Infolan, GES (the former
JVNCNET). These will be joined by many of the RBOCs offering frame
relay, SMDS and ultimately ATM services on top of which TCP/IP runs
(standards and implementations exist and are in use). MCI and AT&T
are also involved in offering such services and outside the US, a number
of start-ups or PTTs are also offering Internet services. There is
a Commercial Internet Exchange, chaired by Mitch Kapor, which forms
a commercial backbone by linking most of the for-profit service providers
in the US.

The federal contribution to the US part of the Internet is probably
on the order of 10% of its total cost (especially if you include the
institutional and corporate investment for LANs). The NSFNET and the
regional networks which it fostered were critical to the expansion
of the system and the NSFNET is still a very important component.
But NSFNET is a service running on the ANSNET which services an
increasing customer base beyond the US Government. The US Government
investment has been absolutely essential for the development and
spread of this technology and it continues to be fundamental to the
continued evolution of the protocols. These funds have been highly
leveraged by industry investments - a partnership we should all be
proud of.

Finally, with regard to the development of networking protocols,
I continue to believe that the "develop, test, standardize" paradigm
is best - make it work first. That's how the Internet Architecture
Board and the Internet Engineering Task Force work. This activity
involves well over a thousand people and is as powerful a technology
transfer vehicle as any I have ever seen. Commercial implementations
of the TCP/IP protocols emerge very quickly out of the standards
process because of the requirement for implementation along with
specification. "Bakeoffs" and "Connectathons" continue to be a
critical part of the culture of the Internet Standards community
and act as a reality check for all would-be standards.

The Internet Society is devoted to the continued evolution and
growth of a multiprotocol computer communications infrastructure
on a world-wide scale. There is room for more than one protocol
in this environment, and that includes TCP/IP.