50 Years Ago, One of the Greatest Inventions Ever?
What was the most significant thing that happened in 1974? Richard Nixon resigning? West Germany winning the World Cup? The IRA bombings in the UK? Well, possibly one of the greatest inventions happened in 1974, and it related to the publication of a research paper [1].
Larry Roberts, Vint Cerf and Bob Khan
On 31 Dec 2018, Larry Roberts died at the age of 81 and left one of the great legacies of anyone who has ever walked on this planet: The Internet. Larry had been a manager in the Pentagon’s Advanced Research Projects Agency (ARPA), and designed the precursor to the Internet: ARPANET. He then oversaw BBN’s implementation of the network in 1969. This involved the creation of IMP (Interface Message Processor), and which were used to route data between the interconnected host networks.
But, how could we scale the IMP towards a structure that would separate the routing of data from transmission and reconstructure of data at a host? The answer came in May 1974 when Vint Cerf and Bob Kahn published a paper for an internetworking protocol and that defined a packet-switched method between nodes:
By Dec 1974, it became an RFC with [here]:
In fact, it was the first document to outline the Internet (and short for internetworking). At the core of the work was the concept of the Transmission Control Program, and which would later become known as TCP (Transmission Control Protocol) and IP (Internet Protocol).
Data segments
The paper is a piece of art and basically outlines the methods of breaking data into data segments, and which are then encapsulated into network packets. Each segment is then given a sequence number. On the receipt of these, a host then reorders them back into the right order. Each segment is then acknowledged for its receipt, and where the sender keeps track of segments that have been successfully received. If there is no acknowledgement for a segement, the sender will then resend it. If this is not acknowledged, the sender will then time out the connection.
As a host could be communicating with many other hosts, Vint and Robert defined the concept of a Source and Destination port number. Thus, different connections with the same host can then be identified with a different port identifier:
But, what about when a sender is sending data too fast for the receiver to buffer and process? Well, Vint and Robert defined a window method, and where the sender and the receiver agree of a window size for the segments. The window then defines the maximum number of segments that can be sent before the sender must stop and wait for an acknowledgement.
Flow control
The control of the flow of messages is then defined using a number of flags:
The SYN flag identifies the start of a connection, and ES and the end of a segment.
The RFCs: TCP and IP
After ARPANET had proven the packet switched networks, the race was now on to scale the principles of it into a large-scale network. For this, DARPA adopted the methods defined by Vint, and, in September 1981, they published the classic Transmission Control Protocol (TCP) RFC [2]:
and and also IP [3]:
By 2004, Vint and Robert were awarded the Turing Award for their foundation work. And, so, at the age of 81, Vint is still highly active and is the vice president and Chief Internet Evangelist at Google.
BBN
In fact, we would probably not have the Internet without one magical little company (BBN) and the vision of Larry Roberts. At the time, most had the word “FAILURE” written over the ARPANET project, and if it had failed, the Internet would probably never happen.
If we go right back to the creation of ARPANET, it was Larry Roberts who published an RFQ (Request For Quote) to interested companies. The task was to build an IMP (Interface Message Processor) and route data across an interconnected network, and this connect disparate computer systems together. While most things at the time focused on cumbersome and centralised circuit-switching, Larry wanted to use a packet-switched approach.
And, so, the big companies prepared their bids and did their usual tendering processing — and basically took what they had, would just deliver to the requirements. Few of them had any faith in what was being built and could only see this as another failed government research project that went nowhere. And to integrate with academia, too, was always going to be a challenge, as academics would want to build something that protected their resources while enabling them to extend their research. In fact, IBM’s solution was to use the large System 360 mainframe computer to undertake the task of routing data.
But, while companies like DEC, Raytheon and IBM failed to see how the creation of the IMP would go anywhere, there was one company that put its heart and soul into the bid: BBN. In fact, it is thought that they spent around six months of time developing the bid. For this, they did a full investigation into the working of the IMP, and had even investigated the hardware and code that it would require. And, so, while they were honest in saying that it was going to be a major challenge, they then laid out the route to the solution and shared their insights. This showed to Larry that, like him, this was not just another project but one that would match the vision of the company.
And, for such a project, most of the companies defined long chains of authority and management, whereas BBN’s approach was to have a single point of focus, and a simplified management approach. Basically, there was a single contact for every question, rather than long lines of delegated responsibility.
At, the time, people used to say, “No one gets fired by buying IBM”, so Larry was laying his whole reputation on the line by going with this small company, which had little in the way of resources to compete with IBM or DEC. But, they had passion and vision and wanted the contract with all their lives. The company were successful in other ways and did not need the grant to sustain them- but they knew its importance. A failure of this project, and there would be no more building of packet-switched network — and possibly no future Internet. And, so, they invested much more time than virtually all the bidders put together.
In fact, BBN were actually first to have an Autonomous System Number (AS1). This is a special number which makes routing on the Internet so much easier, as we just need to know which autonomous system to give our data too, in order to get it routed to the destination. This can be an intermediatory route through the AS, or where the AS hosts the target device.
The choice of an AS approach — using BGP (Border Gateway Protocol) — has really been one of the most fundamental elements in building the Internet at scale. While not perfect, it works! BBN also strived to secure BGP, as it was fundamentally important that no single entity — especially a malicious one — would take over the routing of the Internet. In fact, BBN invented the link-state routing method, and which allowed the “best” route to be discovered to a destination — through the intercommunication of routing tables from devices.
Now, Level 3 Communications uses AS1.
BBN, too, were one of the first companies to be an internet service provider and were the second organisation in the world to register a domain name (on 24 April 1985 with bbn.com):
Domain Name: bbn.com
Registry Domain ID: 4240240_DOMAIN_COM-VRSN
Registrar WHOIS Server: whois.corsearch.domains
Registrar URL:
Updated Date: 2023-05-09T19:30:10Z
Creation Date: 1985-04-24T05:00:00Z
Registrar Registration Expiration Date: 2024-04-25T04:00:00Z
Registrar: Corsearch Domains LLC
Registrar IANA ID: 642
Registrar Abuse Contact Email: domains-abuse@corsearch.com
Registrar Abuse Contact Phone: +1.8007327241
Domain Status: clientTransferProhibited https://icann.org/epp#clientTransferProhibitedAnd, here is the BBN Web page from 1985 [here]:
Why were BBN so successful? They had a passion and a drive, and they recruited the best talent around. In fact, BBN was sometimes know as the “the third university” in Cambridge, alongside Harvard and MIT. The key to any innovative company’s success is their HR function, and in making sure they get the best talent around. One great engineer with a passion and drive can often trump teams of hundreds. But, they must want to do the work — so must be given stimulating and challenging roles.
After innovating in so many areas, in 2009, BBN became a wholly owned subsidiary of one of the companies they beat off for the ARPANET contract: Raytheon.
Conclusion
If you read the paper, you will see that it mapped out the key principles of TCP, and led to the creation of an open system interconnection (OSI) model. A key advancement was the separation of the network routing part and the data segment element. At its core was the principle of uniquely identifying a session and then identifying each segment with a segment identifier. This identifier can then be used for flow control, and where the sender knows how many segments have been received correctly.
Reference
[1] Cerf, V., & Kahn, R. (1974). A protocol for packet network intercommunication. IEEE Transactions on communications, 22(5), 637–648.
