A review of the origins of the internet

Cast of hundreds

Most of the material for this article is from wonderful a book “Where wizards stay up late: the origins of the internet” by Kate Hafner and Matthew Lyon

There are many people and institutions involved in constructing the ARPAnet which later became the Internet.  The book “Where wizards stay up late: the origins of the internet” has been well researched and makes for a great read.

In this article I highlight some key characters and points that stick in my mind.  There is a great deal more credit due to many characters.

Bob's problem and ARPA

The story starts with General Bob Taylor around 1966.  Bob's problem is simple to explain, however I suspect it will be near impossible for young digital natives to grasp.

Next to Bob's Pentagon office sat a computer room with three computer terminals.  One connected to a computer at MIT (Massachusetts Institute of Technology), one connected to University of California in Berkeley and one connected to Strategic Air Command in Santa Monica, California.  Bob needed one computer terminal for each computer he wanted to work on.  If Bob wanted to switch from messaging Berkley students about surfing monster waves, to planning the deployment of nuclear weapons with Strategic Air Command he had to physically move from one computer terminal to another.

This problem can be broken into two key elements.

First, to have a computer terminal connected to a computer required a continuous physical wire connection.  Each of these computer terminals had a long distance copper wire connecting the terminal to the computer.

Second, one brand of computer could not communicate with another.  So an IBM machine was not at all compatible with Honeywell or DEC.

Bob was the boss of Advanced Research Projects Agency (ARPA).  ARPA was a bit short of things to do after NASA stole their thunder by taking over space flight, so Bob and ARPA set about creating the ARPAnet which would later become the internet.

 

Problem 1: Line Switching Vs Packet Switching

Line Switching

Until the first message was sent on the ARPAnet on 29 Oct 1969, all wired communication used line switching.  To place a phone call or connect to a remote computer required a physical connection between the two end points.  The physical connection could only be used by one phone call or computer connection at a time. 

The technology is called line switching.  Early telephone systems required a human operator to manually create a physical connection by plugging one line into another.  Later, electromagnetic and electronic systems replaced human operators. 

Even with electronic switching, the limitations of Line Switching still exist:

1) A physical connection is required between the two communicating parties

2) A physical line can only carry one phone call or computer connection at a time.

3) If the line breaks at any point, the connection is lost.

 

Packet Switching

Packet Switching was invented by two people, Paul Baran and Donald Davies around 1966. 

Paul Baran worked in the US and his motivations for packet switching were centered around making communication more robust (nuclear war resistant) for military purposes.

At the same time, unknown to each other, Donald Davies was working on a similar project in the UK.  Donald's motivations were to improve communications at a fundamental level.

The concept of packet switching is simple enough.  Divide a message into small packets of data.  Each packet has a header which includes the source and destination of the message.  All the sender needs to do is pass the packet to the nearest node on the network.  The packet knows where it is going and finds its way though the network until it reaches its destination.  A packet only uses the physical line for the duration of its transmission, after which the line is freed up for other users.  Packet switching solves the three key problems of line switching:

1) A physical connection is not required between the two parties.  Each party (sender and recipient) only requires a connection to the nearest node on the network.

2) Physical lines can carry many messages at a time.

3) The network has redundancy.  If one connection on the network is broken, the packet can find another route.

As an analogy to compare line switching with packet switching, imagine sending a physical letter.  The line switching approach is paying a taxi driver to hand deliver the letter to the recipient.  While the taxi is delivering your letter, she is not able to deliver any other messages.  The taxi approach may be the fastest, most direct way to send a message, however it is not very efficient.  The packet switching approach is to use the postal service.  With the postal service, you put a message in an envelope (the packet body) and write the address on the outside (the packet header).  You then post the letter at your nearest post box (node on the postal network).  The letter is passed around inside the mail system until it pops out at the address you specified on the envelope.  The postal service can handle many other messages at the same time as delivering your message.

The importance of the invention of packet switching cannot be overstated.  Today, the vast majority of electronic communications, wired and wireless, use packet switching technology to reliably carry enormous quantities of data for multitudes of users over great distances.

AT & T

In the 1960s AT&T was massive.  Prior to being broken up by the government in 1982, AT&T enjoyed a monopoly over the telephone network.  AT&Ts approval was required to do anything on the phone network.  No organization could connect a device without AT&Ts permission.

Paul Baran tried hard to promote his packet switching concept to AT&T in a series of seminars with AT&T engineers.  AT&T said packet switching could not be done, and their own line switching technology was the only way to communicate. AT&T wanted nothing to do with packet switching or the ARPAnet project.

 

Problem 2: Computers are not compatible

In the 1960s the issue of computer incompatibility was far greater than today.  Today it is annoying when some websites and applications look a little different on different devices.  In the 1960s computer manufactures such as IBM, DEC, Honeywell and the like had completely different computing hardware and software.  There was no common USB style port to connect machines to devices.  They spoke completely different machine languages.  There was no intention of the machines being able to communicate with each other. 

The prospect of getting a dozen different manufactures to make their computers talk to each other was daunting, bordering on impossible. 

The Interface Message Processor (IMP)

A computer scientist named Wes Clark proposed the idea to “Leave the hosts alone”.  Use a separate computer (the IMP) to do the communication.  The host computer would only need to learn to talk to the IMP, rather than figure out how to talk to many other devices.

The IMP was an early version of the modern day router.  The first IMP was constructed from a large fridge sized Honeywell DDP-516.  Today's routers, particularly those inside mobile phones, can be as small as a thumbnail sized microchip.

Again, the importance of this single discovery cannot be overstated.  All computers and devices which communicate with the internet use some type of router rather than communicating directly. 

IBM

IBM was the dominant player in the computer market at the time.  When ARPA sent out requests for proposals to build the first IMPs, IBM declined the opportunity saying it was impossible to build a computer cheap enough to act as a router. 

 

The Opportunity lost

This is pure speculation, 100% hind-site on my part.

Cisco systems are the number one internet router company today, with a Market Cap of 208 Billion USD.  There are many other companies around the world making routing equipment, however Cisco usually stands out as number one.

If AT&T and/or IBM had got involved with the ARPAnet project when they were invited, those two companies would be supplying much of the worlds internet routing equipment, making it impossible for a company like Cisco to exist today.  So arguably AT&T / IBMs arrogance in the 1960's cost them at least $208 Billion in Market Capitalization today!