The Internet is, of course, a global packet
switched data network and the early history of its development
centres mainly on the evolotion of packet switching technology -
work on this subject happened independently in the UK and in USA
and there was also a certain degree of sharing of ideas...
In England, at the NPL (National
Physical Laboratory) Teddington site, Dr Donald
Davies conceived the idea of a "network" of inter-connected
data terminals where the data was broken into small chunks (or
"packets" as Davies named them) rather than in a continuous
stream, which also avoided the problem of short messages being
blocked behind long messages
Davies' ideas were first presented in public
in the USA at the ACM
symposium, in Gatlinburg in 1967, and in the UK at the IFIP
Congress, 1968, in Edinburgh
The NPL packet switching approach was adopted by the US Department
of Defense in 1967 and in 1968 aproject called the ARPANET, the
forerunner of the internet, was launched
The first ARPANET link was between the University of California
and Stanford Research Institute in 1969 and the first international ARPANET connection was made between
London and Norway in 1973
In the USA efforts to interconnect computers had been going on for
some time but had focussed on circuit switching rather than packet
switching
In 1959, a computer scientist, Paul Baran,
was working at the RAND
Corporation on the idea of interconnecting computers. In
1962, Baran presented a paper titled On
Distributed Communications Networks he proposed the computer
communication concept of standard message blocks routed as "hot
potatoes" in a store-and-forward system
In 1965, two computer scientists Dr.
Lawrence G. Roberts and Thomas Marill, conducted an
experiment to understand what it would take to interconnect two
computers, namely a TX-2 computer at MIT Lincoln Lab with a Q-32
computer at System Development Corporation in Santa Monica CA,
using a lease-line from Western Union - this experiment
highlighted the complexity of the problem and concluded that
circuit switching, as used by the telephone network, was a poor
solution
Roger
Scantlebury of NPL presented at the aforementioned
Gatlinburg conference in 1967 on the local
network being developed at NPL (National Physical
Laboratory) in the UK by Donald Davies, which used much
higher speed circuits - he also saw the US plans for
ARPANET and reported back "It would appear then that the ideas
in the NPL paper at the moment are more advanced than any
proposed in the USA"
Dr Roberts was heavily influenced by the NPL work, stating
later "The NPL paper clearly impacted the ARPANET in several ways.
The name "packet" was adopted, a much higher speed was selected
(50 Kilobit/second vs 2.4 Kilobit/second) for internode lines to
reduce delay and generally the NPL analysis helped confirm the
concept of packet switching
Here is a video of Larry Roberts
(Dr.
Lawrence G. Roberts) himself talking about much of the
history described above, including talking to Donald Davies of
NPL in the UK, about using the term "packets" and faster links
(at 11:10)
The key point about the approaches taken in both
England and America was that processing power almost dictated that
seperate computer systems were used to handle the establishment of
communcations between the main computers
In the video on the right,
Leonard Kleinrock, talks about what was first probably
the message sent on a computer network on Oct. 29th, 1969, between
UCLA and Stanford University
The word "login"
was typed but the system crashed and only "lo" was received
The first IMP had just been installed at UCLA in
Kleinrock's lab on that day in October 1969 and the first ARPANET
communication was established between with Douglas Engelbart's lab
at the Stanford Research Institute (SRI), as the commemorated by
the UCLA Office of Public Information below
By December 1969, the fledgling ARPANET had grown to
4 nodes, with the connection of the University of California at
Santa Barbara and the University of Utah
In 1972 the second phase of ARPANET
produced a major expansion from the 4 original sites to 40 sites
across the USA
In 1972 ARPA was renamed the DARPA (Defense Advanced Research
Projects Agency) in 1972 - there is an interesting section in RFC 1000,
which summarises RFCs 1 to 999, which starts:
The procurement of the ARPANET was initiated in the
summer of 1968 -- Remember Vietnam, flower children, etc?
There had been prior experiments at various ARPA sites to link
together computer systems, but this was the first version to
explore packet-switching on a grand scale. ("ARPA" didn't
become "DARPA" until 1972.)
In 1973 University
College of London in England and the Royal Radar Establishment
in Norway were connect to ARPANET
In 1975 satellite links to Hawaii and UK were added and there
were already 57 nodes in the network and the 1st of July
- That year the Defense
Communication Agency (DCA) took over direct control of
ARPANET, as there was risk to national security because of rapid
growth and complete lack of control over access and activity
An electronic mail box protocol was
also discussed in various RFCs from RFC 196 to RF720 from 1971 and in 1976 Queen Elizabeth II made history
by sending an electronic mail announcing that the Royal Signals
and Radar Establishment in Malvern was available on the ARPANET
and by this time the familar format of "name@host"
was in already in us
The video on the
left, celebrating the 50th Anniversary of ARPANET shows its
evolution during the 1970s and the interntetworking between
various elements, which is also covered below
Robert Kahn
of BBN also explains the network topology and resilience, with the
aid of diagrams drawn on a whiteboard - he is very probably the
industry's very first SE (Systems or Sales Engineer) - he goes on
to detail such familar topics as store & forward devices,
network management, remote admin and code upgrades
As Robert Kahn explains, the IMP
(Interface
Message Processors) used to form the ARPANET were
also suplemented by TIPs (Terminal Interface Processors), which
allowed connection of simpler terminals and were therfore early
terminal servers
Peter
T. Kirstein in England is often recognised as the father of
the European Internet as he was responsible for the first ARPANET
connected system outside of America
The British side of the story is remembered by members of the
original NPL (National Physical Laboratory) team in this video
The comparison with visiting ARPANET people from
America is mentioned and the connection of the NPL network with it
- most likely the first example of InterNetworking
Even an example of a data network carrying encoded
speech is explained - probably the first example of VOIP in the
1970s
This diagram of the ARPANET protocols shows a
multi-layered approach, much like the OSI 7 layer model and TCP/IP
5 layer model, which it obviously predates by many years, albeit
with only 3 layers
IMPs are interconnected at the first layer with
physical circuits and communicate via packets
Hosts are interconnected at the second layer with
virtual links and commicate via messages
Users processes are interconnected at the third
layer with virtual connections and communicate via byte streams
The ARPANET host-to-host communication used 1822 protocol, as
covered in the specification document on the left, along with
hardware and interfacing details
Data messages contained the destination host's address and the
data message being sent and the 1822 hardware interface allowed
the IMP to deliver the message to that destination address, either
to a locally connected host, or to a downstream IMP the last IMP
in the path transmitted a Ready for Next Message (RFNM)
acknowledgement to the sending host IMP.
Messages had a total length of 8159 bits, the first 96 bits were
reserved for the "leader", or header, and the remainder was data
1822 messages had guaranteed delivery and if a message did fail
to be delivered, the IMP sent to the originating host a message
informing it as such - this system proved not to be foolproof
though
Later versions of the 1822 protocol, such as 1822L, are described
in RFC 802 and
its successors
Stephen D.
Crocker, a graduate student at UCLA, created and led the
Network Working Group (NWG) and wrote the very first Network
Working Group Request for Comment
(RFC) which summaried the IMP Software, the Host-to-Host
Software and the Host Software - Click the image on the right for
the original copy
RFC 2 explained how to write RFCs and added that the Network
Working Group "seems to consist of Steve Carr of Utah, Jeff
Rulifson and Bill Duvall at SRI, and Steve Crocker and Gerard
Deloche at UCLA. Membership is not closed"
The problem with unreliable message delivery on the ARPANET was
addressed with the Network Control Program (NCP), which provided a
standard method to establish reliable, flow-controlled,
bidirectional communications links among different processes in
different host computers.
The NCP interface allowed application software to connect across
the ARPANET by implementing higher-level communication protocols,
an early example of the protocol layering concept later
incorporated in the OSI model
NCP was developed under the leadership of Stephen D.
Crocker, then a graduate student at UCLA. Crocker created
and led the Network
Working Group (NWG) which was made up of a collection of graduate
students at universities and research laboratories sponsored by
ARPA
to carry out the development of the ARPANET and the software for
the host computers that supported applications
RFC 384
lists all the OFFICIAL SITE IDENTS FOR ORGANIZATIONS IN THE ARPA
NETWORK in August 1972- it only amounts
to 4 pages
The ARPANET Completion Report on the right describes much of the
history covered above and also has more diagrams showing its growth
in the glossary at the end
In 1980 the whole ARPANET came to a complete halt because of a
message timestamp issue, which mimicked a DDoS attack - the details
were recorded in RFC
789 - at that time ARPANET had 213 hosts, with a new host
added approximately once every 20 days
The NCP protocol had some critical limitations:
Inability to address hosts and networks further downstream
than a destination IMP on the ARPANET
Lack of end-to-end host error control, meaning packet loss
causing protcol and and application crashes
Error control was never deemed necessary for the ARPANET as it was
designed as the only
network in existence
As changes in the ARPANET were ruled out, it was clear that new
protocol to replace NCP would be required - the new protocol would
need to operate more like a communications protocol, rather than a
device driver orientated approach that NCP took
In 1974 Vinton
Cerf and Bob
Kahn wrote the paper on the left detailing a new protocol
called TCP (Transmission Control Protocol), to correct all of the
shortcomings of NCP by itself
This early TCP implementation worked well with file transfer and
remote login but problems emerged with packet voice
experimentation in the 1970s, which showed that packet loss would
be better corrected by the application, outside of TCP
The answer was to split all of the functions of the original TCP
into two protocols and in 1978 TCP was split into:
A simple internetworking protocol to provide only packet
addressing and forwarding
An advanced protocol providing flow control and packet loss
recovery
These protocols, namely Internet Protocol (IP) andTransmission
Control Protocol (TCP) were combined into a protocol suite,
commonly known as TCP/IP for ARPANET
By 1983 the DCA (Defense Communication Agency) and DARPA had
made the historic move to establish these two new protocols as
de facto for the ARPANET - The Department of Defence (DoD)
also adopted TCP/IP and the paper on the right by Vint Cerf
and Edward Cain describes the implementation within the DoD
Peter
T. Kirstein was instrumental in defining and
implementing TCP/IP alongside Vint Cerf and Bob Kahn
Vint Cerf himself describes his
work and collaboration with all others invloved in this
video, including the TCP and TCP/IP protocols He also covers need for interworking, stemming from
the decision to allow seperate networks, with and
"Gateways"between them, which would later be renamed
"Routers"
There was a period in the industry of what has been termed the Protocol Wars,
which saw great debate over which protocols and standards should be
used for computer networks and the Internet
IBM had previously produced the Systems
Network Architecture (SNA) in 1974 and Digital Equipment
Coropration (DEC) had released DECnet in 1975,
which both competed with TCP/IP
The OSI model is a conceptual 7 layer framework for network
communications which is still referred to for the Internet, LANs and
WANs
The OSI and the 7 Layer Model actually lost the Protocol Wars
though and TCP/IP became the de facto choice for networks small
and large worldwide
TCP/IP defines 4 layers, as above, but it
practice 5 layers are involved when the physical layer (copper
cabling, optic fobre, wireless radio etc.) is included
As can be seen above, there is no distiction bewteen the OSI
layers 5, 6 and 7 (Session, Presentation and Application) and they
are all considered to be in OSI layer 7 (Application) on all
current IP networks, including the Internet
This doesn't seem to stop industry professionals
from trying to place various applications into the unused layers 5
and 6 - It's interesting that IP Load Balancers are called Layer
4-7 devices but they are not really layer 5 or 6 devices
The need for internetworking is illustrated well by the
diagram on the right, which shows how ARPANET and
MILNET, the part of the network that carried
unclassified United States Department of Defense
traffic, was physically separated in 1983.
The ARPANET served the academic research community
Gateways relayed
electronic mail between the two networks
Other direct connectivity between the networks
was disconnected
SATNET, or the Atlantic Packet Satellite Network as it was also
known, was an early satellite network that was
an important part of the first heterogeneous computer network
and the fledgling Internet
As the diagrams on the left and below show, SATNET was an
international link across the Atlantic Ocean to England and
mainland Europe, which used the Intelsat IV
and Intelsat
V geostationary communication satellite
Goonhilly, in Helston,
Cornwall provided the satellite link from the UK and, surprisingly,
a mobile van from Stanford Research Institute, as part of PRNET (Packet Radio
Network)
The SRI van looked much like a bread van but was actually a full
ARPANET node, equipped with a DEC LSI-11, packet radio kit, a
shielded generator and air conditioning
In 1976 on the 27th of August, the van was parked next to a
well-known Portola Valley, California biker bar called Rossotti's
(now the Alpine Inn),
with cables running to a picnic tables and it was then that first
two-network TCP/IP transmission was conducted between the van and
ARPANET
The first demonstration, linking SATNET, the ARPANET, and PRNET took place on
the 22nd of November 1977, when data flowed through the mobile SRI
van at Menlo Park, California and the University of Southern
California in Los Angeles via London, England, across three
different types of network: packet radio, satellite, and the
ARPANET, thus demonstarting the first Internet transmission between
three disparate networks.
From 1977 to 1978, the SRI van was also used for the first VOIP
(Voice over IP) tests, using ARPA's Network Speech Compression
Program, over the Mickey Mouse phone in the van
The video on the left Robert Kahn talks about various parts of
ARPANET, including IMPs, packet radio and the invention of
gateways, which eventually became "routers" in the Internet
The report on the left details testing conducted on SATNET in 1988
for TCP throughput perfomance, round trip, packet loss etc. as well
as its structure
By November of 1983 the size of the ARPA Internet
was causing huge problems with maintaining records of IP
addresses, which was done essentially manually, with hosts being
entered in a global table, the HOSTS.TXT file, on
the Network Information Center (NIC) at SRI - this approach was
not able to scale any further though and this manual booking was
at the limit of its capabilities and a distributed database, rater
than a centralised one, was needed
.
A working group including Jon Postel
(who hither to had done much of the maual address admin), Paul
Mockapetris and Craig Partrige was set up to solve the issue
and they duly published RFC 882
which furthered the work on name servers in RFC 819 and
created the
This naming service meant that end sers could now use host names,
such as USC-ISIF, instead of IP addresses, such as 10.2.0.52
The very first DNS server was called "Jeeves" and was brought
online in 1983-84, when the Internet had something over 1000
hosts, by Paul Mockapetris on DEC Tops-20 machines at the
University of Southern California’s Information Sciences Institute
(USC-ISI) and SRI International’s Network Information Center
(SRI-NIC) - in 1984 RFC 920
was published and the Domain Names System (DNS) was named and the
full service came into operation
Using a grant from DARPA, another DNS server was written in 1986
on the Unix operating system by graduate students at the
University of California at Berkeley - this was the Berkeley Internet Name
Domain (BIND)
Jon Postel passed away, his obituary was wriiten by Vint Cerf and
published as RFC
2468 entitled "I REMEMBER IANA", a reference to to IANA the Internet Assigned Numbers Authority,
it rememebers John as the first person to carry out these services
personally, it begins "A long time ago, in a network, far far
away, a great adventure took place!"
Local Area Network (LAN) deployment gathered pace in the 1980s and
proprietary mail systems, such Unix
Mail gained usage, along with X.400 based systems
Until the early 1990s, it looked like these systems of a part
of the Government Open Systems Interconnection Profile (GOSIP)
would become the de facto system of electronic mail
The adoption of TCP/IP on the Internet saw Simple
Mail Transfer Protocol (SMTP) protocol implemented on the
ARPANET in 1983
NSFNET
began with 56-kbps links, which were upgraded to T1 (1.544 Mbps)
links in 1989.
The growing size of the Internet started to pose
issues for the selection of routes taken by data packets in end to
end flows across the entirety of its reach, as depicted below right,
and also the best protcol for sharing this routing information (at
Layer 3 of the TCP/IP model) -
Routing at this early phase of the Internet was carried out by just
a few Gateways (later to be called Routers at the networks core -
these held tables of complete Layer 3 network reachability of the
entire Internet
The core Internet routers exchanged information with Gateway-to-Gateway
Protocol (GGP), which was descibed by RFC 823 on
the DARPA Internet Gateway, written by BBN and published in 1982
GGP used a distance vector algorithm, based on the Bellman-Ford
Algorithm, for best route selection, which only uses a count
of the number of gateways/routers to traverse, or hops, as they are
called, with out any information on the speed or latency/delay on
the interconnecting links - none of the gateways or links were very
fast at this time though, so the hop count worked as a viable metric
RFC 827 ,
also written by BBN in 1982 as a draft, described the Exterior
Gateway Protocol EGP and states "In the future, the
internet is expected to evolve into a set of separate domains or
"autonomous systems" - RFC 904
Exterior Gateway Protocol Formal Specification was published in
1984, as the formal specification for EGP as the protocol for
reachability information exchange between Autonomous Systems of
gateways
EGP still used a simple distance vector algoritm and only catered
for only a single path between any two end points on the
network, with its tree based network topology - this was of little
consequence at the time though, with NSFNET acting as the single
“backbone” handling all long distance traffic - EGP furthered the
definition of “Autonomous Systems” by introdudicng AS Numbers
for each individual network running EGP
RFC 1009,
Requirements for Internet Gateways, R. Braden, J. Postel (June 1987)
- Autonomous Systems IGP EGP ARP
* Local-Area Networks (LANs)
LANs may have a variety of designs, typically based upon
buss, ring, or star topologies. In general, a LAN will
cover a small geographical area (e.g., a single building or
plant site) and provide high bandwidth with low delays
*
Wide-Area Networks (WANs)
Geographically-dispersed hosts and LANs are interconnected
by wide-area networks, also called long-haul networks.
These networks may have a complex internal structure of
lines and packet-routers (typified by ARPANET), or they may
be as simple as point-to-point lines
Click the image above for the Patent of the
Transceiver
Click the image on the right for Bob Metcalfe's memo on the Aloha
Network and his ideas for the ETHER NETWORK
NSFNET was further
upgraded to T3 (44.736 Mbps) rates by the end of 1991 and
the NSF also funded a number of regional Internet service
providers (ISPs) to provide local connection points for
educational institutions and NSF-funded sites.
the Commercial
Internet eXchange (CIX) was
In 1991 the NSF allowed Advanced Network and Services (ANS), a
non-profit company established by the Merit Network, IBM, and
MCI to carry commercial traffic over the ANSNet backbone, the
same infrastructure that carried traffic for the NSFNET
Backbone Service
In mid-1991, meetings that led to the formation of the CIX
were held in Reston, Virginia. The original signatories to the
CIX agreement were PSINet, UUNET, and CERFnet.[3][4]
While testing was originally done in the Washington, DC area,
commercial operations began at a PSInet facility in Santa
Clara, California in the Fall of 1991.[4] In April 1996, the
CIX router moved to a more neutral site in Palo Alto,
California, the Palo Alto Internet eXchange.[4]
the NSFNET, was originally intended as a backbone for other
networks, not as an interconnection mechanism for individual
systems. Furthermore, the "Appropriate Use Policy" defined by
the NSF limited traffic to non-commercial use. The NSFNET
continued to grow and provide connectivity between both
NSF-funded and non-NSF regional networks, eventually becoming
the backbone that we know today as the Internet. Although
early NSFNET applications were largely multiprotocol in
nature, TCP/IP was employed for interconnectivity (with the
ultimate goal of migration to a standardized Open Systems
Interconnection [OSI] set of standards — that never appeared).
In 1993, the NSF decided that it did not want to be in the
business of running and funding networks, but wanted instead
to go back to the funding of research in the areas of
supercomputing and high-speed communications. In addition,
there was increased pressure to commercialize the Internet; in
1989, a trial gateway connected MCI, CompuServe, and Internet
mail services, and commercial users were now finding out about
all of the capabilities of the Internet that once belonged
exclusively to academic and hard-core users! In 1991, the
Commercial Internet Exchange (CIX) Association was formed by
General Atomics, Performance Systems International (PSI), and
UUNET Technologies to promote and provide a commercial
Internet backbone service. Nevertheless, there remained
intense pressure from non-NSF ISPs to open the network to all
users.
1994: BGP-4
1991WAIS, invented by Brewster
Kahle, is released by Thinking Machines Corporation.Gopher is introduced by Paul
Lindner and Mark P. McCahill from the University of Minnesota.
1993 – WWW invented
World-Wide
Web (WWW) is released by CERN in Geneva, Switzerland.British researcher, Tim
Berner-Lee creates HTMLThe web as we know it is born!
The growth in the number of websites on the Internet has been truly
phenomenal, of course, with the total now
but from 1991 to 1992, the first year, the totsl only grew to 10
sites and only to 130 by 1993 - the exponential growth then
continued, as the chart below lwft shows
My journey with computer netorks started in the eralry 80s when
worked on early LAN systems from Altos Computer Systems based on
RS422 and ealry hardware based S-Net star networks from Novell - click
here for details
By the mid 80s I often worked on UUCP file sharing but in 1986 I
installed TCP/IP at the London Stock Exchange (only around 2 or 3
years after it was adopted by ARPANET) - click
here for the details and I went on to teach Ethernet and
TCP/IP to engineers - click
here for more
As the annotations say on the graph, I wrote my first homepage in
pure HTML in 1994, when they were only around 3,000 sites on the
Internet and only a mere handful of personal/business sites
Click here to view my homepage
in 1994, with links to the then existing websites now
preserved on the WayBackMachine, the Internet Archive
I also wrote a fully costed Business
Plan to encourage the business to move into structured
cabling, 10BaseT networks and Network Computing - click on the
image on the right to view
I brought up my first fully online webpage in 1995 on a small
computer under my desk at my offcie in Slough, UK and I proved
this to be true by accessing it from a machine in California at
the HQ of Bay
Networks (slow to load), while I was there on a networks
training conference that year, which was the first time I saw a
page from England appear on a screen in USA.. the Internet was
truly an international phenomena!
Sorry but there's only pictures to look
at here so far - I'll add some words when I get time!!
Until then though, this TCP/IP
Bible may prove useful! - Here's some more detailed info on
OSPF and BGP .....
plus a Design Guide
for large IP networks (covering OSPF & BGP)