The first instance of long distance transmission of digital data is unrecorded as probably it was a message sent on a tom-tom. In more modern times, it is interesting to note that digital transmission preceded voice transmission by about thirty years. This, of course, was by means of the telegraph.
Communications between data processing machines also has a longer history than many of us would realize. The earliest instance I know of was at the Wright Patterson Airforce Base in the early 1940’s. This was a simple process of converting punch card information to paper tape, transmitting the latter and then reconverting into cards. However, the intensive use of data communication involving computers is really a product of the 1960’s and this discussion will be about what is available now and where the field could be going.
A definition of data communications is useful to help divide the whole communications area into categories. Mr. V.W. Wolontis, Executive Director of the Data Systems Engineering Division at Bell Labs (Bell Laboratories Record, August 197 0) noted that data communications is really anything that is not either voice transmission or video transmission. This is an important distinction because obviously any of these three categories could be either digital or analogue and the important distinction is the end use, i.e., it is not the primary purpose of transmitting either voice or video to alter the content of the information other than perhaps photo enhancement in the latter case.
Essentially then, I will be talking about data communications although I will later try to draw all three categories together into a meaningful pattern.
BACKGROUND FOR DATA COMMUNICATION
Most of us have had some exposure to data communication through the use of a time sharing terminal or similar device. We have all had experience with a telephone and will realize that the latter is essentially an analogue device. We know then that in order to transmit data over a phone line through the switched network we must have a device called a modem which changes the pulses coming from the computer into tones that are then sent over the network to another modem which reconverts these tones into discrete impulses which the computer can utilize.
As this process takes place over exactly the same switching network as is used for voice communication, the common carriers are understandably concerned about network pollution, i.e., some device not manufactured by the common carriers suppliers which might inject tones that would cause the switching network to malfunction. We should remember that the tones are of the same type as those created by dialing the telephone. You can hear these clearly when you place a call on a Touch-Tone phone. This is the background to the controversy over the use of foreign attachments such as acoustic couplers. The industry argues that such devices should be allowed as long as they meet the necessary specifications for the network. This battle has more or less been won by the computer industry as a result of the now famous Carterphone case.
It is worth, for a moment, considering what happens on a switched network as this is rather fundamental to understanding some of the problems that can arise. The earliest form of switching data was the torn tape method. In this method, if you wanted to send information between two points on a selective basis, you would send data from your terminal to a central location which would be the only termination point for your line. The information would be caught on paper tape which would then be torn off the machine and put on a rack in the centre of the room. The front end of the tape was interpretated with the destination and an operator would manually read this and then go to the terminal connected with the desired destination and feed the paper tape into this. In this system, all transmission links in the network are terminated in a central location.
It is easy to see that the next step would be to substitute for the torn tape system a direct line switching mechanism whereby a mechanical relay operation would read the destination at the front of a message and make the connection between the originating terminal and the desired terminal automatically.
This is exactly what happens when you dial a telephone for a voice transmission and the majority of the switching centres in North America still operate on the mechanical relay basis.
Needless to say, if you dial another terminal which is already in operation, you get a busy signal in the same way you do with a telephone. This and many other problems led to the development of stored program machines as a substitute for the relay systems. Once a line terminates in a computer, the computer can temporarily store the message until the other terminal is ready to accept it. Logically enough, this process is known as a Store and Forward System. At this point then, systems for data communication start to diverge somewhat from the straight voice communication. While it is possible for an Electronic Switching System (ESS) to register the fact that a normal telephone call was unable to get the line desired and then ring that line as soon as it is free and establish the connection, a computer cannot readily store the voice message to be relayed when the connection can be made. It should be noted this is far from impossible but is not planned for the immediate future.
This ability to store and forward is one of the advantages of a digital network compared to one that was designed primarily for vocal transmission. However, we should be aware that if the transmission is essentially from computer to computer, this same storage of data can be handled at the intelligent computer terminal rather than at the central switching system.
I might add that there are many other reasons for wanting a pure digital system as opposed to one that is designed for voice.
One reason is the elimination of the need for an expensive modem. The error rate is another consideration. To quote Mr. Wolontis “typically for the great majority of switching offices, Data-Phone service gives you less than one error in 100,000 bits on about 90% of the connections dialed”. If 100K bits equals approximately 12,500 bytes and the average print line is 120 characters, this would mean that one could expect an error for about every 100 lines printed. Naturally, this does not happen as the one error in 100,000 bits is an average over many dial-up calls and clearly a bad connection can generate hundreds of wrong bits thus biasing the average.
Also, the use of data compression and other standard techniques would really allow you to print many more than 100 lines before expecting an error. However, it is true to say that this rate is far too high for satisfactory data transmission.
One cause of errors when using the present network is that on direct distance dialing, you have no idea how your call will be routed and the longer the routing, usually the higher the potential for error.
Another reason that direct distance dialing may be unacceptable for some computer applications is that the normal delay in making a connection may be as long as 10 to 15 seconds. This would not be adequate for such things as airline reservations systems.
This leads us then to a consideration of the second method for data transmission which is the private or leased line. These lines may be standard 2400 baud circuits which are the equivalent to your telephone line, or they may have higher rates of transmission. These lines can be ‘conditioned’ to improve their data transmission capabilities. Because these lines are essentially hard-wired, one can better control the quality of the transmission. If one were to transmit between Halifax and Montreal, this would likely take place over a private line to the common carrier centre in Halifax. The message would then be relayed over the microwave network to Montreal where once again a line would be available for your exclusive use from the termination point of the microwave network to the receiving terminal.
Switched networks are available for up to 48 00 baud or the equivalent of two voice grade lines. The recently announced Multicom service, available from the Trans-Canada Telephone System, could provide high-speed switched capability up to 50,00 0 baud.
THE RELATIONSHIP TO THE COMPUTER
At this point it is worth trying to relate the capability of the communications network to the capability of the computer. As the terms ‘low-speed’ and ‘high-speed’ data transmission are not well defined, let me suggest that we call low-speed transmission that which is essentially oriented toward the speed of a human operator. This would be the type of service that would be used with Telex or TWX or with the time sharing systems in the classical sense of the word. These speeds range up to about 200 bits per second.
I said time sharing in the classical sense because there is always some confusion about just what is meant by this phrase. I will use the term time sharing in the sense of Time Slicing, i.e., where each of many low-speed terminals are given an essentially fixed slice of the time available on the main computer and I will contrast this with Remote Batch Entry where there is usually little interaction between the human operator and the computer and jobs will normally take as much time in the computer as is required to run them asynchronously. This distinction is important when we consider what one wants to do with the line capabilities available.
For example, it is clear that when someone is working at a key driven terminal developing a FORTRAN program, he uses very little of the capacity of either the line or the machine. Regrettably, except for Telex and TWX, there are no low-speed switched facilities available. This means that many users of a key driven terminal are forced to use a standard 2400 baud voice grade line when, in fact, only about a tenth or less of this capacity is required. This has led to many attempts to concentrate many users on one voice grade line.
For example, it might be possible for a time sharing service company to have a device in Halifax which could accept ten or more low-speed terminals and use one voice grade line for relaying this information to Toronto. This increases the costs for equipment but does decrease the line costs. In the United States AT&T does offer Datrex which is a concentrator service not yet available in Canada. Needless to say, there are many who believe that the expense of providing this type of service should be borne by the common carriers and not directly through purchase by the users.
For remote batch processing, the normal terminal is one that can read a deck of cards and print at speeds ranging up to 300 lines a minute. A 2400 baud line can just sustain such speeds under optimum circumstances.
Consider that a 2400 baud voice grade line can transmit the equivalent of 300 8-bit bytes a second. If you were to use such a line to transmit a reel of tape, this tape would likely have been written at 60,000-320,000 bytes per second. In other words, the line capability would be slower than the rate at which the data was written by more than 3 orders of magnitude!
As pointed out earlier, a great deal of work has been done to concentrate data by eliminating blanks, using coding techniques for strings of like characters, etc., but regardless of the techniques used, the limitation of the line speed is what normally limits the speed of a remote batch operation. However, a 4800 baud line with data compression can operate a terminal satisfactorily at around 300-500 lines-per-minute.
It is interesting now to consider the cost to a user of this type of operation. If we were to analyze a typical RBE customer, we would find that he might be spending about $5,000 per month. If the customer were located in Toronto, it would likely be that he would spend about 20% of this amount for the cost of the line and the modems.
Another 20% or about $1,000 would be for the terminal, leaving about $3,00 0 as the amount expended for computer time. As the line, modem, and terminal charges are non-productive to the user, this means that only 6 0% of the money he expends goes directly toward productive use. It is natural, then, that companies involved in supplying data services are constantly trying to reduce the cost of the non-productive part of the service.
We should now take a look at the capacity of a large computer relative to the line speeds we have been discussing. If I may use the example of our own System/360 Model 85, it is interesting to note that the channel capacity on this equipment is now over 36,000,000 bits per second and this will be increased to 60,000,000 bits in mid-1972. Naturally not all this capacity is planned for data communications use as much of it will be involved in the internal processing involving tapes, disks and drums but this further illustrates the disparity between the capability of the lines in the network and the capability of a large computer.
WHAT IS REQUIRED?
Some years ago, AT&T made the prediction that the load for data communications would be equal to voice traffic sometime in the 1970’s. This comment has led to a number of proposals for the immediate establishment of large digital networks. One of these proposals was that of the Science Council recommending the Trans-Canada Computer Network (TCCN). This would be a digital network in addition to the facilities already available for analogue transmission. The analogy is drawn to the importance of this for Canada similar to the building of the railroads in the mid-18 00’s.
In the United States, similar large digital networks are being proposed by private groups such as DATRAN. Some of these proposals have come from the common carriers and some from private organizations. The common carriers have fought the entry of private groups into this field and a breakthrough came with the decision in the United States to allow Microwave Communication Inc to establish a link between Chicago and St. Louis. For an example of small industry versus the large monopoly I would recommend you read an article entitled “The Battle for Data Communication” by Charles J. Lynch in Innovation No. Eleven, 1970.
The costs to the user for lines and modems is large as pointed out earlier. To put this in prospective, at SDL the forecasted annual expenditure for lines and modems is over $11,000,000 by 1974. This is roughly equivalent to the initial investment in the System/360 Model 85. This makes it very tempting for firms such as SDL to look at going into the data transmission business. However, this is properly the business of the common carriers who are naturally concerned that such an approach to the business would ‘cream skim’ i.e., would likely service only the attractive part of the market leaving the common carriers to service the rest. If these and other concerns are foremost in the minds of the common carriers, I believe they should take the initiative now to assist in reducing the costs of communications to the industry.
The common carriers have their own problems in raising capital and the thought of building a massive new digital transmission network may well cause them to hesitate. Much of the problem for the next few years could be solved if the common carriers would consider re-tariffing their present service and would undertake to develop better and lower cost modems.
I could probably be convinced that there is a front-end loading on a switched network to set up a call, but I find it almost impossible to believe that it really costs substantially more once the microwave network is established to transmit data from Halifax to Vancouver than it does from Halifax to Toronto and yet the charges are now on an almost linear charge/distance relationship. A smoothly decreasing non-linear charge/duration relationship for switched inter-city service or dedicated service would go a long way to solving many of the problems we now face.
Also, before we build a new network when there is clearly more than adequate capacity for the next number of years on the existing networks for long distance transmission, we should reconsider whether the original statement by the AT&T that data traffic would equal voice traffic, is in fact, true. In the September/October issue of Computer Magazine, the Stanford Research Institute analyzed just where the load on the network would really come from.
Their analysis indicated that by far the largest load by 1990 would remain the telephone requirement. Next to this would be Video Telephone or Picturephone. These would be followed by some form of electronic mail and then television.
The largest computer oriented operation would be remote library browsing, presumably as an adjunct to some computer assisted instruction operation.
This does not necessarily mean that we should not look at building a new digital network because as I mentioned earlier, any of the above requirements can be met either digitally or through analogue transmission. However, we should be careful not to waste our resources when a relatively simple solution, such as the rate changes suggested might serve the users in the short run. Another example of more efficient use of present facilities would be economical pricing and packaging of the lightly used night shift on the microwave network for overnight tape transmission.
In any case, one thing is clear and that is that the common carriers have a huge job to meet the very heavy development costs required both for the short run needs of the computer industry and for their own long run needs to meet the massively increasing communication requirements of our country and of every other country.
THE WIRED CITY
To give you an example of the incredible investment required one need only consider what will have to be done if even some of the 2 8 suggested requirements for the communication network proposed by Roger W. Hough of the Stanford Research Institute in fact come about by 1990. Many of these items are directed toward the individual user in his own household. If one realizes the amount of information that will be transmitted to and from each household, it becomes obvious that our cities would have to be completely rewired. At the moment community antenna television (CATV) does bring a substantial capacity into nearly a million households in Canada. This capacity could be enough to transmit 25 to 3 0 TV channels. However, this is one-way transmission and is not suitable for things such as in-the-home instruction. Also, this system is not switchable.
Ideally, a system of switched coaxial cables would solve the problem if the cables were capable of about 300 megahertz. It should be noted that a TV signal requires about 6 MHz and 600 telephone signals could be accommodated in the same space.
Needless to say, the switching of lines of such capacity would, in itself, be quite a problem (remember the channel capacity of a large computer) but in fact, this would not be required for many of the services.
It is this broad area that is now being addressed by the new Canadian Computer/Communications Agency and the implications for the correct use of capital in our country, in the coming decades are enormous.