This Session
ARPANET as precusor to Internet
Origins of the Internet Exploration of heritage from this
Origins of TCP/IP and Internet
How Internet defeated rivals
I202, Fall 2003 Comparison to Open Source Philosophy Session 11 Thomas Haigh (Web and commercialization coming next time…)
i202, Session 10, Thomas Haigh 1 i202, Session 10, Thomas Haigh 2
Origins as ARPANET Early Network
Funded by DOD research group ARPA Advanced Handrawn diagram of Research Projects Agency APRANET in 1969 Built by BBN, an MIT spin-off firm Four nodes (“hosts”) Each using different kind of ARPA funded several computer science teams computer for graphics, HCI, AI, etc. Three in CA, plus Utah Wanted them to share expensive resources Cross country to MIT Also wanted to test new networking ideas added in 1970 Original function is for remote logins 111 hosts by 1977 Like Telnet today Non ARPA sponsored sites pay up to $100K to connect
i202, Session 10, Thomas Haigh 3 i202, Session 10, Thomas Haigh 4
Actual Applications Shift Key Features Preserved
See Abbate readings on this Packet Switching Official resources (other computers) Hard to find (no directory) RFC standards process Hard to use (no documentation) Hard to get account on (no incentives or accounting) Separation of Net from Applications New applications created over existing infrastructure Interconnection of different kinds of IMP as LAN hub for local machines Email is the surprise hit computer Initially runs on top of file transfer protocol Science Fiction reviews list & Project Gutenberg
i202, Session 10, Thomas Haigh 5 i202, Session 10, Thomas Haigh 6
1 Interconnection of Separate Net from Applications Different Kinds of Computer
Previous networking tended to be multiple dumb Code to handle network terminals to one computer traffic is compartmentalized No need for application Or computer/computer link, coded from scratch programmer to worry about details ARPANet includes many different kinds of Network code written and computer debugged once only As purpose is to give remote access to ARPA funded Originally done via hardware research systems, save duplication Interface Message Processor or IMP Makes standards more important Dedicated minicomputer delivered to all installations Today done in software Net code built into OS i202, Session 10, Thomas Haigh 7 i202, Session 10, Thomas Haigh 8
Packet Switching RFC Process
ARPANet was first network to use Still basis of modern Internet Request For Comment Doesn’t establish a circuit for Used since 1969 on ARPANet duration of connection Become official record of net’s Unlike telephone Design decisions Standards Architecture Anyone, in theory, can submit proposal Contributions are edited and numbered Breaks message into many small “packets” Currently RFC 3638 is newest Each one addressed and numbered Each one passed across network separately Provide “public” access to protocol specs Reassembled when they arrive at destination And chance for net community to comment on proposed “Dumb network, smart clients” changes
i202, Session 10, Thomas Haigh 9 i202, Session 10, Thomas Haigh 10
TCP/IP TCP/IP: Advantages
Current foundation of Internet Designed by Vince Kahn and Bob Cerf Simple and reliable Two protocols: Works with different kinds of physical media IP = Internet Protocol (packet addressing & transmission) Satellites, ethernet, modems, radio, etc. TCP = Transmission Control Protocol (retransmission, etc.) These are “under” TCP/IP protocol Became standard on ARPANet around 1982 Can add new media without changing code Design and testing from 1975 onward Works with many different applications Open standard – anyone can use DOD backs for military use Protocols for news, ftp, etc. run “over” TCP/IP Many different firms produce equipment for Addition of web, streaming video, etc. needs no changes to TCP/IP or to network
i202, Session 10, Thomas Haigh 11 i202, Session 10, Thomas Haigh 12
2 The Internet Was Spread of Internet
Name given to collection of INTERconnected TCP/IP NETworks Grew rapidly ARPANet (closed in 1990) 1984: passes 1,000 hosts 1990: 313,000 hosts on 2,063 networks NSFNet for scientific use (2000: passes 100 million hosts) MILNET for military use Opened to commercial use in 1991 (NSFNet) Other US and international networks Until about 1994, remained mostly for science, Bridges constructed to other kinds of networks government, academic use Allows interconnection of networks using other But became truly global system protocols for internal traffic “Gateways” to on-line services, FIDONet, etc. Success of TCP/IP means most corporate networks, Increasingly popular with students, hobbyists universities, etc. now use internally as well
i202, Session 10, Thomas Haigh 13 i202, Session 10, Thomas Haigh 14
Other Kinds of Networks Rival Philosophy
In 1980s, no such thing as “internet service OSI project of International Standards provider” Organization Access to academics and researchers only Need account on connected computer Sets up “seven layer model” (see paper) So most network users are elsewhere Encourages separation of different aspects of Other academic networks networking, like TCP/IP but more so E.g. BITNET for IBM sites Commercial on-line services Designs specific protocols for standard use Aimed at individuals and small business X.25 for packet switching (instead of TCP/IP) Amateur bulletin boards X.400 for email delivery FIDONet links them together for news, email
i202, Session 10, Thomas Haigh 15 i202, Session 10, Thomas Haigh 16
Internet has Similarities Lots of OSI support to Open Source
Through early 1990s, seen as future Lack of concern with commercial issues Standards for protocols reflect long process Don’t address concerns of commercial users Interests of telecommunication firms Standards and information freely available Hardware suppliers Actual code tends to be shared too Try to address Start with basic system to solve immediate problems Long term needs of Gradually expand as needed Very wide range of users Fix problems as they arise Design philosophy is flexible, modular So tend to be Complicated Users are also developers Inflexible In many cases, at least Tied to specific applications and assumptions
i202, Session 10, Thomas Haigh 17 i202, Session 10, Thomas Haigh 18
3 Downsides of Net Philosophy Example 2: Spam Internet mail relied on trust, small user pool When technology widely used, can be Basically, can specify any name, organization, hard to change email address and return address
Might have been better to plan more Mail is relayed to destination without charge No mechanisms to authenticate sender Example 1: Running out of IP addresses Or prevent tampering, or encode, or recall TCP/IP v4 limits number of address numbers message, or acknowledge receipt, or guarantee (4 billion sounds like a lot, but given out in delivery blocks) Breaks down in modern world Can add some of these features to email clients TCP/IP v6 would fix, but adoption is slow But would require mass simultaneous adoption Might also involve recentralization, loss of privacy
i202, Session 10, Thomas Haigh 19 i202, Session 10, Thomas Haigh 20
4