Licklider’s Intergalactic Computer Network - The Computing Universe: A Journey through a Revolution - page 212

called Cyclades. By 1975, the TCP specification had advanced sufficiently to
be implemented concurrently at three sites - BBN, Cerf's group at Stanford,
and a group headed by Peter Kirstein at University College London. In October
1977, the effort reached a significant milestone when Cerf and Kahn demon-
strated sending messages across three interworking networks: a packet radio
network, a packet satellite network called SATNET, and the ARPANET.
The networking protocol took its final form in early 1978 when the part of
the transmission protocol that deals with the routing of the packets was sepa-
rated off as a discrete Internet Protocol (IP). Under the new TCP/IP protocol, TCP
was now responsible for breaking up messages into datagrams, reassembling
them at the other end, detecting errors, and resending lost messages. The IP
was responsible for routing the individual datagrams. Jon Postel ( B.10.17 ) later
summarized the guiding principle for what should be in the IP: “I remember
having a general guideline about what went into IP versus what was in TCP. The
rule was 'Do the gateways need this information in order to move the packet?'
If not, then that information does not go into IP.” 32
In the first chapter we mentioned the layered approach as one of the fre-
quently occurring “mental tools” used in computer science. The TCP/IP protocol
is a further illustration of this idea (see Fig. 10.17 ). TCP/IP consists of four layers.
At the top there is the application that sends a message to another application
running on a remote computer. At the bottom lies the actual physical wire or
fiber where the message is translated into electric or light impulses. If the mes-
sage is long, then it is sliced up into smaller pieces, each put into a separate
envelope. Imagine if we sent an entire topic to somebody one page at a time.
On the Internet, as the message travels down it is put into a bigger envelope at
each layer. We can think of the protocol as resembling matryoshkas , or Russian
nesting dolls ( Fig. 10.18 ). On the receiving end, the message travels upward,
and at each layer the envelope is stripped off. Because the original message was
sliced into pieces, the message must also be reassembled on the receiving end.
TCP/IP emerged as the dominant Internet standard after a long battle with
a rival proposal for standardizing networking. The rival standard was called
Open Systems Interconnection (OSI). It had been developed by the International
Organization for Standardization (ISO), a world body that tries to establish uni-
form sizes and other specifications to ease the international exchange of goods.
The U.S. government and European national governments all decreed that OSI
was the official networking standard for the Internet. Similarly, major com-
puter manufacturers like IBM, Digital, and Hewlett-Packard also adopted the
OSI standard instead of TCP/IP. Amazingly, it was the popularity of the Unix
operating system and Ethernet LANs in universities that turned the tide in
favor of TCP/IP. Unix, developed at Bell Labs in the 1960s, is a popular oper-
ating system because it is powerful and stable, and can be installed on many
different kinds of machines. Except for Microsoft Windows, nearly all major
operating systems have some kind of Unix at their core. Bill Joy, later one of
the founders of Sun Microsystems, received an ARPA grant to write the TCP/IP
stack , a complete set of networking protocols, into the free Berkeley version of
Unix. The first Sun machines were sold with Berkeley Unix including the TCP/
IP networking software. With the worldwide popularity of Sun workstations
and the increasing availability of Ethernet as a commercial product in the early

Application layer

Transport layer

Internet layer

Network access layer

Fig. 10.17. The layered structure of the
TCP/IP protocol.

Fig. 10.18. Matryoshkas are Russian
nesting dolls. Here they illustrate the
concept of message encapsulation.

B.10.17. Jon Postel (1943-98), “the
unsung hero of networking,” was
the coordinator of the RFC dis-
cussion forum and later became
chairman of the Internet Assigned
Numbers Authority, in charge of
assigning Internet numbers. Cerf
wrote Postel's obituary, which was
published as RFC 2468.