Roberts also predicts the eventual replacement of telephone leased lines with a variety of more efficient alternatives. He says as traffic grows, present trunklines, which work at 56 kbps, will be replaced with T1 digital trunks which operate at 1.544 megabits per second, allowing a great increase in the transmitting capacity of the system.

He also says satellite transmission will become economically feasible for long-

Contmues to page 56

NET as a research tool and to access the research sites in the community," Roberts said. "One of the goals we had was to lower the cost of computer resources, and we did that by a factor of three by allowing universities to share resources instead of duplicating them."

ARPANET also quickly dispelled the worries of the communications engineers who had said it would not work. It did not lose packets; it did operate smoothly and efficiently.

Ma Bell's Reaction

ARPANET did not impress everyone, however. One group that was not excited at all was American Telephone and Telegraph, the parent company of Bell Telephone, as Roberts found out when ARPA tried to sell the network to them.

"The Bell lab's evaluation said it wasn't compatible with anything else in the Bell network," Roberts said. "Actually it wasn't incompatible with their network— we were using rented Bell lines—it was incompatible with their thinking."

Bell apparently had decided back in the early 1960s that the Sherman Anti-Trust legislation made it impossible for them to get involved in data networks, so they never did any research in them. They have lived to regret that decision.

Today they are struggling to develop their own packet network under the name Advanced Communications Systems (ACS). Roberts says they are still putting hundreds of man-hours into designing the system and are at least two years away from building it. Meanwhile both Telenet and Tymnet, Cupertino, CA, are operating smoothly.

One way or another digital data transfer will take over, probably in this decade.

"For the home market it will take longer," Roberts said. "In business it is happening very quickly, as businesses go

11Actually it wasn't incompatible with (Bell's) network— we were using rented Bell lines— it was incompatible with their thinking,"

to leased lines connecting buildings full of terminals to the network directly."

The problem with the home market is the microcomputers are spread out. To solve this, GTE, the Japanese and several other groups are working on digital telephoning. In fact, the Japanese telephone company is presently installing a small operational digital telephone system.

GTE is working on a method to allow 64 kilobit (kb) data transfer through paired wire, an idea which Roberts terms quite feasible. He said the system would actually work at about 200 kb, but would be alternating so you could transmit in one millisecond, receive in the next.

Roberts envisions a telephone-computer terminal combination of the future which would handle both voice and digital data transfer at the same time, to and from different destinations.

It would, for Instance, tell you while you are taking one call that someone else is calling. It would get the caller's name and number and tell him you are on a call and either put the caller on hold or call back. If he wanted to send a written message, he could enter it and your phone would store it for you.

If you had a list of people you needed to talk to, you could have your phone send a packet to their phones and find out who is free. Your phone would then make voice contact automatically.

Your phone would also be able to take and store messages and communicate with time-sharing computers over the network. It would be able to tell callers if you are unavailable and when you will become available. It will be able to direct them to another phone where you can be reached.

Roberts said the cost of lines makes it absolutely necessary the two systems, voice and data, be combined, although it is not impossible that voice transmission would remain analog for some time and the two methods would simply share the same cables.

Roberts is working in the forefront of this development with GTE. He is responsible for the development and marketing of hardware that will complete the digital communications revolution.

He remains optimistic about the future of packet switching. "It's my belief that the only way to handle data is to do the economics and the protocol conversions (to allow different machine to communicate with each other)," he said. "It will continue to be the major and probably the only \$/ay to handle data. There's really no other way to do it." ■

by Bert Latamore 80 Microcomputing staff

Ethernet Alive at Stanford U.

Local networking may be the hottest thing in business communications since the telephone. A network changes the microcomputer on an executive's desk from an isolated business machine into an integral part of a company-wide computer system.

It allows an executive in his office in one building to instantly send written information, charts, diagrams, worksheets, etc., to another executive in another office in another building.

"Ethernet is one of 20 to 30, .systems

It allows him to compose a letter, correct it himself, then send it directly to a spinwriter in the mailroom for printing. It allows him to receive up-to-the-second information from the firm's central data banks on any aspect of his company's activities from parts in stock to the book balances whenever he needs them.

Such a tool creates a large business

Networks continued trom page 55

distance transmissions when the excess data flow exceeds 100 packets per second between two cities. Because of the built-in 270 ms delay in one-way satellite transmissions this would not be acceptable for transmission of interactive traffic but could be used for batch data transmission.

Roberts also spoke of packet radio with local distribution systems. Local distribution is the most expensive part of a network. In this alternative to land lines, several clients would share a frequency. Each would broadcast to the node only when it actually has data to send, he says.

Tymnet recently announced it was trying a packet radio system in San Francisco and a satellite link between San Francisco and New York City. The experiments follow a January FCC ruling reallocating the 10.55 to 10.68 GHz RE spectrum to digital electronic message services, both for local distribution microwave transmis-

market, and many firms, from large computer manufacturers like Digital Electronics Corp (DEC) to tiny components makers, have developed networks and are supporting them with the necessary hardware and software.


One of the most interesting of these is Ethernet. Several manufacturers, including Xerox, DEC and Intel have promised to provide support to it.

This network, developed by Xerox, is barely out of the experimental stage, and most components are not yet available for it. An experimental version is running at Stanford University, Stanford, CA.

The Stanford Ethernet is a three megabit system, and the equipment and software used were part of the experimental system Xerox built when developing the 10-megabit Ethernet. Xerox donated the experimental equipment to Stanford, Car-negy Mellon University and MIT. A group at Stanford has been working with the experimental system. It is operating successfully, interconnecting 25 computers on campus. The object of the experiment sions and satellite communications.

As the cost of copper continues to increase and the cost of computing drops, it becomes easy to visualize a future in which dynamic allocation networks dominate telecommunications. Cable television (and radio), with its promise of unlimited channels, is growing rapidly. Packet switching could greatly improve its economics.

Even telephone service may go this way. However, Roberts says, voice transmissions cannot be condensed as much as digital transmissions can, so such a change may not be economical in the near future, especially as it would involve the complete rebuilding of a large in-place system. Because of this, electronic mail sent through packet networks may become serious competition for the telephone.

Whatever the future, the networks are here today which provide a doorway to a whole new world.■

by Bert La tarn ore 80 Microcomputing staff is to determine whether Ethernet or Ethernet-type network technology has a place on Stanford's campus.

10,000 Computers Expected

The current population on campus of 250 computers and around 2,000 terminals is expected to grow to something like 10,000 in 5 to 10 years. They hope to develop a network or system of networks to allow most or all of these to interconnect. While a single Ethernet system would not be able to connect all the computers spread out around the several-acre Stanford campus, it could be used on a more limited scope in a single building or a large number of Ethernets could be linked together to do the job.

Simple Packet System

Ethernet is a packet system but with a much simpler and more limited method of operation than the intercontinental common-carrier systems. When you send a message through Ethernet, the system chops the message up into packets including addresses indicating what stop on the system the message is being sent to. Then it inserts your packets into the stream of packets running through the Ethernet cable.

So far its operation closely resembles the common network. However, the Ethernet arrangement does not include any packet-switching computers, and all packets follow the same path through a single cable.

Each client computer on the system reads the addresses of every packet going past. If the packet is for it, the computer captures and stores it. If not, it passes it back into the system.

Ethernet by itself is even more limited than that. Ethernet is not a complete system; it provides only the bottom two of the seven levels of "protocols" or software and hardware packages needed for a total network system. What it does do is give you the connection devices to attach your machines to a cable and the interfacing software to allow it to create and address the packets.

It does not include application programs, for instance, which would instruct machines on how to communicate with each other. Ethernet is meant to'provide a common network base over which the networking of different firms' machines can be accomplished.

Four Hundred Databases Await Your Microcomputer

This does not mean that DEC machines will be able to talk with Xerox computers—unlike the common carrier networks, Ethernet does not handle any interfacing problems. These are taken care of by the higher level protocols which the Ethernet package does not include.

Ethernet also does not include any provision for confirming that the receiving machine gets the message sent to it. If that machine is turned off or malfunctioning or if one of the packets are addressed incorrectly, the message will never arrive, and without some kind of confirmation the sender will not know this. This must also be included in the higher level protocols.

Physical Limits

Ethernet has physical limits as well. It is a baseband system as opposed to a broadband system. This is like a telephone line which is a baseband system able to handle comparatively few messages at once while a television cable is a broadband system able to carry hundreds of channels simultaneously.

Additionally a single system cannot be more than 2,500 meters long. You can connect several Ethernets together using "gateways," however. These are devices used to interface different networks. Alternatively, an Ethernet system can be connected to a different kind of network, including the common carrier networks, using gateways which would translate between Ethernet and the other network's protocols.

Not On Market

Part of the problem with determining the economics of the system is that the commercial hardware and software is not yet on the market and not expected to be for 18 to 24 months.

Meanwhile, Ethernet is only one of some 20 to 30 possible networking systems, each with their own advantages and disadvantages. Some, like the Tymnet local network which is really a smaller version of its common carrier network complete with packet switches, have much greater capability in terms of area covered and number of machines it can handle.

Others, like Omninet by Corvus Systems, which does not even use coaxial cable but runs on simple twisted pair wire, have less capability than Ethernet but are probably less expensive.

The question is what system fits the individual's needs and pocketbook. ■

by Bert Latamore 80 Microcomputing staff

Compuserve, The Source and, recently, the Dow Jones Electronic News, have received a great deal of press attention as interactive electronic information services for general microcomputer use.

They are, however, only the start of the list of available data bases. Several hundred of these exist in 35 or more countries.

Most were not originally designed for use by home microcomputers; however, the networks automatically handle most compatability problems, and as the microcomputer market grows, more data base publishers are becoming conscious of compatibility and are making efforts to take care of formatting and other residual problems.

It is impossible to list all of these services here. Both Telenet, Vienna, VA, and Tymnet, Cupertino, CA, publish directories showing most of the services they connect to. If you want to subscribe to a service, you must deal directly with that service, not with the network. All network charges are included in the fees you pay for using the service.

Here are a few of the data bases:

• Berkeley Solar Group, Berkeley, CA, offers solar energy analysis, weather data and other services tor the solar energy industry.

• Blaise, British Library, London, England, includes medical data bases, a listing of all British-imprint books acquired by the Library of Congress and a document-ordering service.

• Bridge Data Co., St. Louis, MO, gives common stock quotes and other services to investors.

• Bodart, Inc., Williamsport, PA, offers 800,000 English-language monographs, including a descriptive list and an on-line ordering service.

• Comcap Computing Systems Oakland, CA, provides technical time-sharing services for civil and structural engineers and land surveyors.

• Dr. Dvorkovitz & Assoc., Ormond Beach, FL, offers what is reputedly the largest data base of licensable technology in the world.

• E. H. Boeckh Co., Milwaukee, Wl, provides building cost estimates for insurance appraisers, developers, contractors, etc.

• Florida Computer Inc. (FCI), North Miami, FL, has daily and historical data on the United States, Canadian and London commodities markets for speculators, hedgers and brokerage houses.

• Lockheed Information Systems, Palo Alto, CA, has more than 100 data bases holding more than 30 million records giving references to published articles, conference papers and technical reports on a wide range of subjects.

• National Library of Medicine, Bethesda, MD, provides medical information including a 500,000-citation file of abstracts and references and is one of several medical services available.

• New York Times Information Service, Parsippany, NJ, carries abstracts from The New York Times and more than 60 trade publications.

• QL Systems Limited, Ottawa, Canada, has 40 or more data bases on Canadian subjects, including Canadian law, aquatic environmental data, fisheries, mining technology, energy, business, metals and Parliament.

11Several hunched. databases exist in 35. countries.

• Weather Systems International Corp., Bedford, MA, provides past, present and forecast weather data and past and present sports and business information.

• Westlaw, West Publishing Co., St. Paul, MN, gives full text and headnotes t>f decisions in all federal courts and in courts of all 5Cf states.■

by Bert Latamore 80 Microcomputing staff

0 0

Post a comment