Side_1_360

diverse and dynamic traffic situation of com-

puter networks the enhanced reliability and traf-

fic resilience of mask shaped nets are essential.

A significant part of the development effort was

concentrated on a desire to handle these needs

of computer networking automatically and well.

One issue in that development pertained espe-

cially to mask shaped – as opposed to star

shaped nets. Both traffic control and reassembly

of large messages are more complicated in the

former. But with the resulting TCP/IP (Transport

Control Protocol / Internet Protocol) protocol

suite mask shaped nets can exhibit their advan-

tages automatically in dynamic situations.

The result was the protocol suite TCP/IP. The

successful, persistent and ubiquitous use of

TCP/IP is now established in millions of nets of

even more millions of computers. That deserves

being mentioned explicitly. Those protocols

resulted from an extremely thorough analysis

and design. “No stone was left unturned” during

the development which took several years. The-

oretical analyses were complemented by experi-

ments. Combinations of traffic types and re-

quirements, network topologies and application

types were imagined, implemented, tried, failed,

changed and tried again. The “final” TCP and IP

were not easily postulated and approved.

Traffic Diversity

Nobody can ever reproduce in a laboratory the

chaotic traffic pattern of a lively telecom or

computing network and even less the diverse

demands of information exchange. The growing

active dynamic traffic situation in the Arpanet

prevailed during onwards development of its

own underlying technology. That may be one

reason for the robustness, elegance and surviv-

ability of the result. Arpanet was the laboratory.

At the same time it was an active telecom net-

work, a resource sharing network and a forum of

creative and critical people. During a period of

intensively active development methods were

conceived and perfected until functioning well in

an environment which was closer to reality than

anyone might have dreamt up in a “sterile” labo-

ratory environment. At the same time a profound

theoretical understanding was developed. It kept

its scrutiny on experimental results and was both

guiding and following up the work in an ad-

mirable teamwork. The group at UCLA under

Leonard Kleinrock’s leadership was foremost

in that area.

One of the many experiments illustrates some

of this thoroughness, the Internet conference

speechexperiment. Digital speech coding is of

course a long story in itself. Whereas inter-

national telephony today uses 64,000 bits per

second to represent ordinary speech, many other

methods can represent fully understandable

speech by much more compact codes. One

example is Linear Predictive Code – LPC. It per-

mits understandable speech to be represented by

2,400 bits per second (this is not the limit, but it

was used for that particular experiment). Such

compact coding is less tolerant both of back-

ground acoustical noise around the speaker and

of errors in the transmission channel. Loss of

a packet is more harmful in the more compact

codes. An LPC coder/decoder (codec) developed

by MIT’s Lincoln Laboratory was used for the

experiment. The first version of the codec, a rack

mounted unit so large it could barely be lifted by

a man, was successively replaced by smaller and

lighter units doing the same. It helped clearing

the way towards possibilities of today’s inte-

grated circuit chip solutions.

Further about the conference speech experiment.

A rather many-sided development had created

solutions and understanding of protocol options,

network configurations, packet satellite channel

access algorithms and their inherent stability,

various performance characteristics, and many

other factors, before the final experiment. It

Virtual

connection

Real

connection

Til Ola NormanOrreveien 13B2045 Leirsund

3,50 kr

Host A

• 
  


• 
  


• 
  

  Get file X

  
  


• 
  


• 
  


• 
  
  
  
  • •
  
  
  • 
    
  
  
  
  

Host B

• •


• 
  

Network

X

Virtual

connection

Packet

handling

File transferprogram FTP

Packet

handling

Real connection

Layered protocols let
application programs in
co-operating computers talk
“virtually direct”. Actors
(programs) at each level need
not worry about details of what
goes on at the lower levels of
the information exchange.
An analog situation is in
correspondence by mail,
where users put (and get) the
enveloped letter in a letterbox
and leave the many detailed
procedures (“lower levels of
protocol”) to the various
departments of the Mail