Inside The Internet Notes

Lectures

Intro

• Dawn of Electronic Computing
• Pre-Internet Communication
• Research networks 60s&70x
• The first "Internet" Mid 80s
• The Web early 90s
• Ubiquity of the internet mid 90s on

WWII

• 10,000 people working at Bletchley Park
• BOMBE: Mechanical computer
• Colossus: Electronic computer

Alan Turing and Bletchley Park

Bletchley park is considered the birthplace of modern computing. The day after war was declared several people (including Turing) showed up as had been arranged before the war. First Skunk Works operation. They brought the smartest people in Britain together with a nearly unlimited budget. Bletchly existed primarily to break the Enigma encryption. This was important to intercepting communications from the Blitz Kreig. Polish Mathematicians starting breaking the code. Bletchley built on their work. The polish computer was Bomba, the British was Bombe. Turing built "cribbing" into the Bombe so they could guess some text. Each Bombe simulate 36 enigma machines. There is between 10 and 12 miles of wire in each Bombe. Enigma was only good for short messages up to about 250 characters. The Geheimschreibers - "Secret Writers" were far more complex and could send long messages. The Mark I Colossus was designed to break this computer and went into operation in early 1944. Mark II was delivered before D-Day and was used to confirm that the deception surrounding D-Day worked. The Mark I Colossus used over 8Kw of electricity. Much of the design of Mark I was based of research done for the British Post Office. The machines were left on 24/7 as to not subject the vacuum tubes to thermal shock. The Colossus was rebuilt starting in 1993 by Tony Sale. Turing was not involved in Colossus.

History Through Modems and Terminals

After the war there was a proliferation of electronic computing. Governments valued their mathematicians, scientists and engineers. Communication was initially done over telephone using teletype. Many early systems used tones sent over phone lines.

Leased Lines: Lines leased from telcos for full time access.

Store and Forward Network: Local computers or terminals would talk to a central computer with a leased line that would forward on to other computers in the network. Networks with more hops were slower but cheaper. As a result there was a movement to bring lots of Universities into the network. This was the early 80s, email took 6 hours to 2 days to cross the country. On of the first big networks of this type was Bitnet.

Store and forward had several drawbacks. Queues meant unfair bandwidth usage. One computer being down lead to nodes being unable to reach each-other. It was also very expensive. The question was how to dynamically handle outages and transport many messages simultaneously and efficiently.

By the mid 70's there was a lot of research into packet switching. This lead to the development of routers for traffic control. Routers did not understand the whole network, only local topology. You could potentially have issues where a poorly formed packet ended up in an infinite loop.

Larry Smarr

Started as an astrophysicist. Through the 60's (after Spunik) the government dumped funds into research/academia. In the 70's that was cut - by 78 we were graduating half the number of engineers as in 70. Smarr collected reports on what science could be achieved with access to high power computers. After some computers were established, the telecom industry lobbied to prevent network access from anywhere in the country.

NCSA - Innovation

National Center for Supercomputing Applicaitons at Urbana-Champaign, Illinois.They standardized the TCP/IP protocol. Established in 1985.

Michigan (the state) did not participate initially because they had the MERIT network tying together UofM, MSU and Wayne State.

Doug Van Houweling

Char of the NSFnet which was operated out of the University of Michigan. He's kind of self-important and very boring. He later became the President and CEO of Internet2 - whatever that is. The NSFnet was disbanded because it's purpose "should be filled by private industry".

Reflection on NSFnet

The network was supposed to cost $15M, which was not enough to fund it. UofM came in with an offer (if they received the grant) to throw in an additional $40M. This got the grant (which other schools were not happy with) and let the network be strong enough to be successful. Partnered with Merit Network, IBM, MCI and the State of Michigan. Ran from 88 to 95.