March 29, 2024

TCP/IP and the growth of the Internet

TCP/IP Guidelines

 

ARPANET soon expanded; in Britain and France other packet-switched networks were developed. In 1970, at the University of Hawaii, ARPA funded the construction of a packet radio network, ALOHAnet.

A satellite network was also finding support, both in America and in Europe. It was obvious to ARPA's Larry Roberts that these diverse networks would need to be connected. To help find a solution to this problem, Roberts invited BBN's Robert Kahn to join him.

In 1972, leaders from the various national networks formed the International Network Working Group (INWG). Vint Cerf, a member of the node 1 installation team at UCLA, headed the INWG. A year later, Kahn and Cerf began a collaboration that would resolve the issue of interconnecting networks. They concluded that a "gateway", or routing computer, placed between each of the networks could pass on the transmissions. The gateway would need to communicate with whatever network that connected to it. At that moment in time each network had different rules for governing the transmissions of packets. This issue was further complicated by the conflicting interfaces of these networks. The gateway machines would stand at the interface of two networks and "appear" to each network to be acting as a host computer. ARPANET, with its insistence on utter reliability, demanded that the system provide for similar safeguards. NCP, designed solely for ARPANET, could not provide for this complexity, constructed as it was on the premise of a very reliable network infrastructure.

Cerf and Kahn devised a new set of protocols and presented their proposals in 1974 as "A Protocol for Packet Network Intercommunication". Besides gateways, they introduced the notion of transmissions in the form of a "datagram"; a packet of data. The datagram was to be of specified length, but not fixed in size, with associated delivery information, effectively the sender and destination address. The network would make "best efforts" to deliver that datagram, but it would be the responsibility of the sending and destination host machines to guarantee delivery. TCP would disassemble and reassemble the datagrams in the correct order and deal with errors in transmission. TCP was eventually divided in two; IP handled addressing with TCP dividing and reassembling the data.

The goal was to build a robust, reliable, fast and flexible networking system. It was assumed that networks were inherently unstable. Good failure recovery was imperative if the system was to continue working if, and when, sections of the network went offline. Flexibility was ensured by allowing an existing network to continue operating should new networks be added. It provided a completely reliable end-to-end service. If packets were damaged or missing, TCP/IP would retransmit the message. The protocols were to be non-proprietary; the network architecture would work with any other type of network. Data overheads would be kept to a minimum; simplicity enhanced speed of transmission and packet switching.

By adhering to these guidelines, TCP/IP had far reaching implications for the growth of the Internet with global economic, political and cultural ramifications.

Go here for a hyperlinked illustration of the DoD TCP/IP model.

Next page » The Creation of the Internet

Previous page « The Network Working Group

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Up to top of page