by Mark Ollig
During President Eisenhower’s administration, a data network intended to link together computers all across the country was being designed.
The US Defense Department’s new network under their Advanced Research Projects Agency (ARPA) would be called the ARPA Network, or ARPANET.
The Department of Defense Directive 5105.15, which officially began ARPA, was signed by Secretary of Defense Neil McElroy Feb.7, 1958.
ARPANET would support the wide-area packet switching protocols to be used by computers across the country.
The ARPA network laid the groundwork for the “network of networks,” which we today call the Internet.
It was 35 years ago this week, when this network had the misfortune of crashing.
ARPANET stopped working Oct. 27, 1980, and remained out of service for approximately four hours.
The outage was traced to the data routing tables, which had become corrupted when an Interface Message Processor (IMP) malfunctioned.
An IMP would more commonly be identified today as an Internet router, which is a device used to connect networks together. It also passes data packets of information along the networks.
Further in this column, you will read how the IMP influenced another event.
The cause of the 1980 outage was reportedly diagnosed as a “high priority software process running out of control.”
Available resources were quickly used up in the Interface Message Processors.
The underlining problem was said to be “a rather freakish hardware malfunction . . . causing erroneous controlling data packets to be generated.”
It took more than an hour just to figure out how to restore the network.
Several more hours were needed to locate the faulty-working Interface Message Processors, and get them stabilized so they were operating and communicating with each other properly again over the network.
Of course, many “Internet historians” out there will want to correct yours truly by saying the actual “first crashing” of the early Internet occurred in 1969.
Don’t worry; I’ve got it covered.
The first “host-to-host” messaging from one computer to another, was being tried over the ARPA network Oct. 29, 1969.
The first computer node on the network, said to be the size of a “one-bed-room apartment,” was the Scientific Data Systems (SDS) Sigma-7 32-bit host computer. It was located at the University of California, Los Angeles (UCLA).
This computer node was linked over the ARPA network with the second node, an SDS-940 time-sharing host computer, located over 300 miles away at the Stanford Research Institute (SRI) in Menlo Park, CA.
Using serial communication protocol cabling, the Sigma-7 computer was connected to what looked like a kitchen refrigerator, but was in fact, the previously aforementioned: Interface Message Processor (IMP).
The IMP executed the hold and sending of the data packets being transmitted over the 56kbps modem connected to leased telephone lines.
The telephone lines served as the physical “transport facility” to the second node in this network, the SRI host computer, which was connected to its own an IMP.
The leased telephone lines also functioned as the “data-line network” between the two IMPs.
Charley Kline, a programmer and student at UCLA, was attempting to send the login command, “LOG” to the SRI SDS-940 computer from his user teletype terminal connected to the Sigma-7 computer.
“I would type a character. It would go into my computer. My software would take it, wrap around it all the necessary software to send it to the IMP. The IMP would take it and say, ‘Oh, this is supposed to go up to SRI,’” Kline said during a 2009 National Public Radio interview.
Kline typed the letters “L” and “O,” which the Sigma-7 computer successfully sent to the SRI SDS-940 computer.
Kline was on a headset, talking over the voice telephone line with Bill Duvall, another programmer who was with the SDS-940 computer, at the SRI, in Menlo Park, CA.
Duvall confirmed to Kline the SDS-940 computer received and recognized his typed characters “L” and “O.”
However, a problem occurred when Kline typed in “G.”
The SDS-940 computer did not receive the “G.”
Then, the network went down: it crashed.
A memory buffer at SRI had become full; causing the ARPA network to crash.
Duvall, who was working the connections on the SRI SDS-940 computer, knew why this happened.
One hour went by while he worked to increase the size of a memory buffer.
Kline’s next “L-O-G” command message was sent, and recognized by the SDS-940 computer without incident, (the SDS-940 computer automatically filled-in the “I-N” to complete the “L-O-G-I-N” command).
Kline was now logged in to the SDS-940 computer, and was able to issue commands within it.
The SRI SDS-940 computer in Menlo Park was treating Kline’s UCLA SDS Sigma-7 32-user teletype terminal in Los Angeles, as if it was a directly-connected user terminal.
This proved instead of needing a separate user terminal on your desk with a dedicated connection to every computer you required access, just one user terminal would be able to login or access, any interconnected computer on the ARPANET.
The logbook by Charles Kline shows the very first message sent over the ARPANET was Oct. 29, 1969, at 10:30 p.m.
Kline’s handwritten “IMP log” of this historic message can be seen at: http://tinyurl.com/bytes-log1.
You can view a photo of an Interface Message Processor here: http://tinyurl.com/bytes-IMP.
Here’s a diagram of the ARPANET with the UCLA and SRI nodes: http://tinyurl.com/bytes-arpanet.
Could today’s entire Internet system stop working and crash?
“It could never happen. There are too many redundancies and alternate network routes which prevent the entire Internet from crashing,” many will say.
However, I am reminded about an adage my mom occasionally tells me; “Never say never.”