ATM: The Next Big Thing in Communications
Fast, compatible and becoming affordable
Every business has its Holy Grail. Physicists root through the cosmic strings seeking a Grand Unified Theory of the Universe; empire builders salivate over the perfect hostile takeover. And telecom wizards hunt for the ideal data-transfer protocol: suitable for all data types and all wiring schemes; robust in recovering errors, yet light on overhead; backward compatible, yet a solid foundation for future growth, and so on.
To hear the evangelists of the ATM Forum tell it, at least this Grail has been found. Asynchronous Transfer Mode (ATM) is everything you ever wanted for data transfer, and more. It is scalable over a wide range of speeds and network sizes; it carries voice, motion video and computer files with equal ease. And it’s coming to a telco near you real soon — in some areas, sooner than ISDN.
ATM CHARACTERISTICS
ATM is a group of standards for moving packets of data over a network. It is characterized by several features: switched-circuit topology (think of the telephone system); small packets of data (53 bytes apiece); virtual circuits; and negotiable, guaranteed bandwidth.
Switched-circuit topology. ATM uses a connection-oriented approach. Each device on the network is connected to a central switch, which sets up connections between devices; a connection stays in place until the switch is told to disconnect it. While a connection is in place, a workstation can count on a known amount of bandwidth being available, regardless of the other connections that may be going through the switch. ATM is similar in spirit to the telephone system: local switches connect close neighbors, central exchanges interconnect the switches and fiber backbones span long distances.
This is in marked contrast to computer networking schemes such as Ethernet, Token Ring or LocalTalk. These are all broadcast-oriented approaches; every packet, no matter which station it is addressed to, gets sent to many stations. As you add more stations, the available bandwidth is spread ever-thinner.
The advantages of a switched-circuit network are flexibility and scalability. Each device on an ATM net can run at its own speed; an interface card at the switch handles the clocking and buffering issues. (In contrast, every device on an Ethernet must go at the same speed or the whole net fails.) An ATM network can grow as big as needed without cutting performance simply by adding more switches.
Another advantage is that switches can be connected by multiple paths. If a circuit fails or a fiber is bulldozed, the traffic can be rerouted around the damage. When everything is working, the network can allocate bandwidth to match demand by adjusting the routes.
All packets are created equal. The universal ATM packet (also called a cell) is 53 bytes long — 5 bytes of header and 48 bytes of message — no matter whether it contains voice, video or files. Because all packets are the same size, the network and the switches don’t have to understand each separate kind of data, but can blindly forward packets to their destinations.
Although 48 sounds like a small number, it was deliberately chosen. Audio and video quality suffers if there are long time delays in transmission. (For voice conversations, the threshold of annoyance is about 50 milliseconds of lag time. Anyone who has made a satellite phone call knows how disconcerting longer delays can be.) It is thus important to minimize the time to assemble and forward packets. The shorter the packet, the less time to process it — the packet cannot be sent until the last byte has been digitized.
On the other hand, each packet requires some overhead for network management and addressing. Network efficiency (ratio of overhead to payload) improves and thus transport cost falls with longer packets. Both mathematical queuing theory and field trials indicate that the optimum tradeoff between packetization delay and network overhead is somewhere between 30 and 100 bytes long.
At ISDN (64 kbps) data rates, a 48-byte message incurs a 6-ms packet-assembly delay. There are other system delays, among them the time lag imposed by the speed of light, so you wouldn’t want to start with delays much greater than 10 ms. And with a 5-byte header, this formula yields an overhead-to-payload ratio of less than 11 percent, which is tolerably good.
(As it happens, the choice of 53 bytes owes as much to politics as to math. When ATM was first brought to CCITT, the international telecom standards body, the Europeans had strong technical reasons for proposing a 32-byte message with a 5-byte header. The Americans, for equally sound technical reasons, wanted a 64-byte message plus a 4-byte header. The diplomats, with true Solomonic wisdom, resolved the matter by splitting the difference.)
Virtual addressing. ATM provides two levels of addressing, called virtual paths and virtual channels. Virtual paths are analogous to cables joining user sites. Within the virtual paths are virtual channels, which are analogous to individual wires within a cable. Here, the word “virtual” indicates that we aren’t talking about wires or fibers, but about bandwidth allocations and routing setups along a shared network. The distinction between the levels is made by dividing the address information in the header of each packet into two parts, with 8 bits signifying the path code and 16 bits identifying the channel.
The reason for this distinction is that, with today’s switching technology, there is a noticeable cost to setting up and tearing down a connection. It is often cheaper to keep a connection (say, between campuses of a state university) open for weeks and months at a time than to relinquish the connection only to reopen it a few hours later. A split addressing scheme allows the long-distance switches rapidly to identify packets headed for a common destination, while still supporting the many different phone calls and computer sessions that users may require.
The distinction means that long-haul networks now offer two classes of service: A “permanent virtual circuit” (PVC) or dedicated line between two fixed points, and a “switched virtual circuit” (SVC) that lets you dial up any destination on the net. This year, a PVC is likely to be cheaper — indeed, it may be all that is available. But during the next 18 months we expect SVC service to come down in price, to become ubiquitous and eventually to displace PVC altogether.
Global addressing. One consequence of this path/channel addressing scheme is that there aren’t very many bits available for specifying the packet’s destination. The solution is to separate the creation of an end-to-end connection from the maintenance of each leg of the journey.
Setting up a connection is done with “administrative” packets whose payload contains complete destination and routing codes. As it happens, there is already a worldwide scheme for encoding destinations on public networks: the E.164 standard, otherwise known as telephone numbers. Within local area nets, other schemes such as Internet or Novell Netware addresses might be used. During setup, it is the responsibility of software within each switch to choose the next leg of the route. For this purpose, the switch might query a network name server, consult internal directories and lookup tables, etc.
Once the end-to-end routing has been set up, each switch along the way no longer has to worry about ultimate destinations; it just accepts packets and forwards them to the next intermediate point. For this purpose, a complete address is not necessary. As part of the setup process, each switch in the path would have reported back to the previous switch, either to reject the request (all circuits in use, not enough bandwidth available, switch out of service, etc.) or to accept the request. With each acceptance would come a reference number identifying the path that the receiving switch had prepared for this connection.
This reference number thus acts as a virtual address for packets during one leg of their journey. It will probably be different for each leg. When all is going well, the end user never has to worry about this; he sees a single unbroken link to the destination. But it makes system troubleshooting rather problematic. Existing network management tools (i.e., SNMP) rely on “all points” broadcast queries, which are hard to implement in switched systems.
Negotiated quality. Unlike most other network protocols, ATM contains explicit features for negotiating the quality of service — peak and average bandwidth, maximum allowable packet delay, etc. — that will be available during the call. During the call setup, each switch requests the quality that it wants (i.e., that the user is willing to pay for). In granting such a request, the downstream switch undertakes to protect that quality level for the duration of the call. If it can’t guarantee the requested service, it rejects the request. The upstream switch can try again with a lesser request, seek an alternate path or wait a while.
One novel aspect of quality is the allowable packet-loss rate. ATM explicitly allows a network to fail to deliver a certain (negotiated) percentage of the packets it is given. This sounds terrible, but it actually lets a net work better.
Most communications run in bursts of data interspersed with silence. The network attempts to align the bursts and silences among all its calls so as to deliver the maximum amount of data without excessive delay for any packet. In a well-designed net, most of the time the switch will succeed. Once in a while, though, all the bursts will arrive at the same time, threatening to overwhelm the switch.
Discardable packets. For this reason, ATM contains a mechanism for handling overflows. Each packet header contains a Cell Loss Priority bit; if this bit is set, it tells the switch that this packet can be discarded if there’s not enough network bandwidth to handle it. (As the joke goes, the CLP bit is how a packet with low self-esteem identifies itself.) The situation is analogous to an airline that overbooks flights and allows standby passengers. In both cases, unused capacity (empty seats or idle packet times) is gone forever, so it pays to overbook slightly; but when a rare peak load does occur, the full-price passengers (or the non-discardable packets) get served first.
Discardable packets are especially useful in handling time- and cost-sensitive traffic. For example, when sending a movie down the wire, it is important that the sound and picture stay synchronized, even at the expense of an occasional missing line of video or an occasional blip in the sound. Allowing this controlled degradation of bandwidth means that the movie can be transmitted in a lower-cost channel. On the other hand, some kinds of traffic — computer files, for instance — cannot tolerate any missing packets, but the packets don’t have to arrive on any particular schedule.
WORKING WITH EXISTING MEDIA
An important consideration for any new technology is how well it can coexist with older technologies in the years to come. (Lest you get the idea that old technologies can just be replaced, remember that there are still a lot of rotary-dial phones in use, even though tone dialing has been ubiquitous for 25 years.) For this reason, the ATM Forum has devoted a great deal of attention to making ATM interoperable with existing communication services both long-haul and local.
ATM for LANs. One of the beauties of ATM is that it works equally well for local area networks as for phone systems, cable systems and wide area networks. This means that it will be possible to use the same technology, data packets and protocols for local communication as for remote communication. ATM requires no changes to existing application software. Each computer needs an ATM interface.
ATM works well in local area networks without needing any changes to existing application software. Each computer needs an ATM interface (with appropriate driver software) instead of an Ethernet or Token Ring or FDDI interface. The active or passive hub of the old network is replaced by an ATM switch.
At the moment, a 45-megabit ATM LAN using Class-5 unshielded twisted-pair wiring (UTP5) is roughly as expensive as a 100-megabit FDDI fiber-optic LAN. The cost of one workstation interface card plus a pro-rata share of the switch (or active hub, in the case of FDDI) is about $3,000. ATM cards for optical fiber are now about $300–500 more; fiber switches are similarly more expensive on a per-port basis. Costs are dropping — by 50 percent during the past year — but it will probably be three to five years before ATM costs less than $1,000 per workstation.
Keep in mind that a 45-mbps ATM network can actually have a greater total throughput than a 100-mbps FDDI net. ATM, being a switched system, supports many simultaneous connections, each getting the full 45-mbps bandwidth. FDDI, based on a token-passing ring topology, can pass only a few packets (typically two per ring) simultaneously, no matter how many computers are using the net.
Equally important, existing networks and ATM nets can be logically united by commercially available routers. This means that network planners can phase in ATM as dictated by budgets and demand. The same routers can join older nets to ATM-based wide area networks.
ATM for MANs. For many companies, the most immediate and practical use for ATM will be in metropolitan area networks that interconnect branch offices and outposts with Permanent Virtual Circuits leased from Bell, Sprint, MCI, MFS and others. One 45-mbps ATM line now costs about as much as ten T1 (1½ mbps) lines, so data-hungry customers may actually save money by increasing bandwidth.
ATM around the world. The 53-byte ATM packet format is readily mapped onto most of the existing telco fiber-transmission standards, such as DS3 at 45 megabits per second or Sonet at 155 and 622 mbps. This means that today’s installed base of fibers and exchanges can carry ATM data along with other traffic, and no new infrastructure will be needed. In addition, the telcos have already specified that ATM will form the basis for broadband ISDN.
Equally important, the connections between LANs and WANs are almost seamless. An ATM LAN can simply plug one or more of its ports into an ATM public network. At the interface between the ATM network and other kinds of local-area or wide-area nets, there are commercially available routers from Cisco, Fore and other firms.
The next big thing? Our take is that ATM will rapidly become the new common denominator for high-speed local networks and for long-distance traffic of all kinds. As prices for the gear fall, ATM will take on the stature that Ethernet has today: an expected feature of every decent computer and phone system.
Peter Dyson
A CULTURAL PERSPECTIVE
Let’s ignore the technical differences between Ethernet and ATM for a moment, and look instead at the cultural and sociological aspects. This is particularly apparent in the different methods for allocating bandwidth in the two systems.
[[don't know how to format this table. P. 16 in hard copy]]
Ethernet ATM
Private networks Public networks
Socialist ethos Capitalist ethos
Bandwidth is shared Bandwidth is negotiated
• grab as much as you • ask for what you need
want
• live with what you get • pay for what you get
• based on courtesy • based on contracts
— George Marshall, Adaptive division of Network Equipment Technology
ATM INFORMATION RESOURCES
The ATM Forum, in addition to developing standards and protocols, sponsors educational seminars and presentations that range from layman-oriented to jargon-rich.
ATM Forum
480 San Antonio Road, Suite 100
Mountain View, CA 94040
Phone (415) 962-2585, fax (415) 941-0849
Resources:
Asynchronous Transfer Mode: Bandwidth for the Future by Telco Systems. $10 (plus $5 for UPS shipping; add sales tax in some states) from Advanstar Marketing Services, 7500 Old Oak Blvd., Cleveland, OH 44130; phone (216) 826-2839 or (800) 598-6008.
Understanding ATM. Free from Cabletron Systems sales department; phone (603) 332-9400.
ATM in Private Networking: a Tutorial. Free from Hughes LAN Systems field marketing dept., phone (415) 966-7300, fax (415) 960-3738.
>NEWS
THE WESTERN CABLE-TELCO/COMPUTER SHOW?
Non-traditional cable companies shape the future
In the keynote session that opened the Western Cable show earlier this month, Ted Turner said the past 20 years in the cable business has been a great ride, but it’s over.
The cable industry that Turner, John Malone and others built is doomed. In its place will emerge a new industry whose business is providing two-way interactive communication infrastructure and services. It is an industry that will look as much like the telephone industry as it does the “old” cable industry. It is an industry that may end up funded by, allied with and indistinguishable from the “new” regional telephone industry.
From demonstrations on the exhibition floor to closed-door meetings in hotel suites around Anaheim, CA, the buzz was “interactive technology,” and the influences of both the telecom and computer industries could be felt in almost every discussion.
CABLE’S DIGITAL INFRASTRUCTURE
By far the most impressive thing at the show was the working fiber-optic/coax cable network at CableLabs’ CableNet booth. Anyone still concerned that the cable companies are going to build a digital replica of the current analog system (complete with a finite number of “channels”) would have been delighted by the demonstration.
In essence, CableNet appropriated the best of the latest telephone technologies and married them to the existing coax cable infrastructure to create a new, fully switched, two-way broadband interactive network. (For more on CableNet, see p. 23.)
DIGITAL COMPRESSION
The key to making this viable is the compression/decompression of digitized video. At the Digital World conference in June 1992, Intel chairman Andy Grove argued strongly that digital video compression/decompression would remain expensive and/or of inferior quality for some time to come. The focus for the ’90s would therefore be on using less-than-TV-quality digital video for business applications.
Immediately following Grove’s speech, Dick Green, president of CableLabs, laid out the cable industry’s plans (and time table) for 500-channel digital video service for consumers. To make this happen, the cable industry would have to count on new and/or significantly lower-cost technologies being ready just in time to meet the schedule. The biggest bet would be on video compression technology.
CableLabs bets on MPEG-2. Six months later, CableLabs selected the emerging MPEG-2 digital video compression standard. It appears to have made the right bet. MPEG-2 specifications are now out for final review and MPEG-2 implementations are being shaken down.
At the show, C-Cube Microsystems of Milpitas, CA, demonstrated the state of its MPEG-2 compression/decompression technology. In the demonstration, movie clips were compressed, transmitted and decompressed on the fly using a not-yet-complete MPEG-2 implementation.
The demonstration was running at full 704×480 resolution with an 80-Mbit-per-second transfer rate. There was a total delay of just less than two seconds due to compressing and decompressing the images. C-Cube claims that it will soon be down to something on the order of a 6-frame (1/5 second) delay — plus, of course, whatever transmission delays are encountered. Even the current delay is short enough for live broadcasts such as news and sports. The 1/5-second delay would almost certainly be acceptable for two-way conversations.
We asked C-Cube to run the system for us at various combinations of lower resolution and slower data transfer rates to see what the effect would be. There are definite trade-offs: Lower resolution gives you a “softer” (less sharp) image. Lowering the data transfer rate without decreasing resolution can cause you to lose picture information in action sequences. Most of the systems installed will probably include adaptive logic that will be able to adjust the bit rate on the fly. The system might select a higher bit rate during the action in a basketball game (one of the toughest things to digitize because the camera pans so rapidly), then drop back to a lower bit rate for interviews or commercials.
It can only get better. Things that can be compressed in advance (essentially everything except for live broadcasts) pose a different set of problems. On one hand, extra time (and/or extra computing horsepower) can make it possible to do a better job of compression. On the other hand, system operators will be tempted to move to lower resolution to reduce the amount of digital storage required per movie, and increase the number of simultaneous channels that can be transmitted. In Orlando, for example, Time-Warner will be transmitting at 352×480 resolution and allowing the system to interpolate the extra horizontal resolution. Although this is theoretically as good as VHS tape, to our eyes the images look distinctly “soft” — well shy of the quality of a good analog cable system, let alone that of a good laserdisc.
The systems installed in full-service-network, interactive television field trials in 1994 will certainly do better than the prototype software at the show — and continuing technological advances will make the systems put in place in 1995 and beyond better still. We do not know how good “better” will be, but from what we have seen, we do not think that compression technology will be a serious problem either for tests in 1994 or for more widespread deployment starting in 1995.
FROM MAINFRAMES TO VIDEO SERVERS
Digital video on demand represents a new lease on life for the big computer guys. This is a truly demanding application: billions of bytes of online data storage, extraordinarily high data transfer rates, tens or even hundreds of thousands of online users.
IBM and AT&T, both of which were at the show, along with Oracle and others that were not, are moving rapidly to extend their technology to meet this need. There are no magic breakthroughs here. The technology is not going to be either simple or cheap — and it probably will not be entirely fail-safe either. But the technology should be adequate to the task. And, again, it is going to get better and cheaper during the coming years.
THE SPLINTERED SETTOP BOX MARKET
From a technological standpoint, the settop box, or “user terminal” market, is the area that is still in the greatest chaos.
The traditional suppliers of cable converters, General Instruments and Scientific-Atlanta, were the two early favorites in this area. They had the established relationships with cable companies as well as the technology and the know-how to handle the essential analog cable services, including the extra-cost premium services. However, as phone companies have become more involved, and as the perceived role of the settop box has expanded, other companies have set their sights on what promises to be the biggest personal computer market of all.
Early on, settop boxes were conceived of as very inexpensive personal computers, capable of presenting a truly easy-to-use graphic user interface on the television screen. More recently, there has been increasing desire to equip them with sophisticated graphics processors to support special effects, 3D games and other “sexy” capabilities that go well beyond those available on most personal computers.
Some of the important players jockeying for market advantage include the following.
General Instruments. As you would expect from the market leader, GI has set out to design and build its own box in expectation that it will become the new de facto standard. However, its initial vision was pretty much a garden-variety personal computer without much razzle-dazzle graphic capability.
At the show, however, GI and Intel announced a $1,400 box with a built-in router and Ethernet/LAN output. It will be used in the Castro Valley, CA, trial that Viacom International and Comcast are conducting next year.
Separately, GI unveiled an elaborate settop terminal prototype called the Jerrold LinX module. When inserted into an analog addressable cable box, the module can become a digital interactive terminal. Microsoft, a partner of GI and Intel, is writing a new operating system kernel for the system, which is not based on either Windows or Modular Windows. LinX is expected to be available to consumers in late 1994.
Scientific-Atlanta. A distant number two to GI in the settop box market, SA has followed a different strategy — wrapping its analog know-how and market connections around computers provided by other companies: Silicon Graphics for the Time-Warner Orlando trial, and the 3DO Multiplayer for the US West trial in Omaha.
At the Western show, SA announced a third alternative: its own box based on the Motorola PowerPC RISC processor (the same one to be used in the next-generation Macintosh), plus a new, low-cost graphics chip, called Malibu, which was designed by Kaleida Labs but will be manufactured by Motorola. It is optimized to work with the PowerPC, and has built-in support for the Kaleida ScriptX interactive media programming language. For graphic applications, this box would be more powerful than the PowerPC Macintoshes which will not have the extra graphics chip.
In addition, SA announced an agreement to develop 300,000 boxes for TeleCommunications, Inc. According to SA, the boxes will be customized to meet TCI’s requirements.
As is typical in the current state of settops, none of the SA boxes are compatible with any of the others —let alone compatible with any of the other boxes coming onto the market!
Hewlett-Packard. As a newcomer in the settop box race, computer giant HP garnered a great deal of the media attention at the show, as well as attention from cable operators.
TCI (surprise, surprise) signed a letter of intent to buy as many as 100,000 digital compression terminals from HP during the next two years. This makes HP the first computer company to make the transition into this area of the consumer electronics market. Although both companies were reluctant to discuss any of the details behind the deal, HP representatives said the first generation of the TCI terminal will not support HP’s video printer technology.
HP’s plans are founded on a very hardware-centric approach, according to HP’s Casey Lemas, market development manager for interactive television appliances. In fact, the company does not plan to develop the interface for the TCI-contracted terminals. (According to Electronic Engineering Times, Dec. 6 issue, HP will adopt the Amiga CD32 architecture and employ the Amiga graphics chipset in its first-generation settop model.) HP declined to comment.
“Clearly we are not deluding ourselves into thinking that we are replacing SA or GI,” says HP’s Lemas. “We may be competing with them in this instance, but we might also cooperate in the future. This is not unusual in the computer industry, but it is new to cable.”
Confusion. In addition to the players mentioned above, we cannot discount the behind-the-scenes work going on at AT&T, Zenith, Silicon Graphics, 3DO, Microsoft and Apple Computer — all of which have either directly or indirectly (as in the case of AT&T) stated their intentions to be contenders in the settop box arena. All of this adds up to a mess: We have no idea yet what the user terminal will evolve into, how they will interface with the network, how they will talk to network servers and/or to other user terminals — or even how developers are going to program them. At the moment it looks as if these boxes will all be sufficiently different that any service that wants to exploit the capabilities of the user terminal will have to write different software for each settop terminal to be supported — not the way to create a broad market for interactive services.
There is nothing surprising about this. New markets are usually chaotic. But they do not really start to grow until they become less chaotic.
ONLINE SERVICES DELIVERED OVER CABLE
The user terminal market is not the only area where we are witnessing the marriage of computer and cable technologies. Several of the major cable operators, online service providers and computer companies have announced joint tests to link personal computers to cable TV lines, so that computer users at home can use cable lines to retrieve data as much as 1,000 times faster than over phone lines.
CompuServe. CompuServe, in conjunction with Continental Cablevision and the Teleport Communications Group, is expected to begin delivering online information via cable lines this month. (For more on Teleport, see p. 22.) The pilot program will link Continental Cablevision subscribers in Exeter, NH, to CompuServe Information Services, which has a worldwide subscriber base of 1.5 million. Participants in the cable test will use personal computers linked to the cable lines via Zenith Home Works cable modems. (Zenith has actually been selling cable modems for about 10 years, but until most recently the company has not been able to convince too many people to buy them.)
CompuServe members participating in the test will pay $8.95 per month for unlimited connect-time access to 48 basic services and $8 per hour when accessing CompuServe’s hourly priced service.
Prodigy. In late November, Cox Cable Communications and Prodigy Services announced a similar test using Zenith cable modem technology. According to the companies, Prodigy’s services will initially be delivered to households in Cox’s San Diego system, where the cable plant is two-way capable. No official launch date was announced.
Prodigy and America Online. Not to be outdone, cable operators Comcast Corp. and Viacom International each announced at the show that they plan to test both Prodigy and America Online as part of their respective interactive field trials slated for 1994. Both cable operators have inked deals to use a prototype cable modem under development by General Instruments, in conjunction with Intel. Viacom, of course, plans to provide the online services over cable to participants in its 18-month Castro Valley, CA, field trial, which is expected to be launched in early 1994. Comcast declined to identify the site of its 1994 trial, which it says will last for 12 months.
The cable-PC connection. These announced alliances respond to the growing demand for high-speed data connections for personal computers. Each of the trials provides an excellent work-around to the problem of slow access speeds over phone lines — without demanding a massive capital investment by cable companies.
500 CHANNELS AND NOTHING GOOD ON?
The most disappointing part of the demonstrations held at the show involved the interactive content. Most of what was demonstrated was, naturally enough, either a pretty straight carryover of something already available to computer users via online services or a displacement of something that already exists (i.e., movie rental).
But why should anyone be surprised or concerned? Most people have always supposed that the initial applications will be the obvious: movies on demand, online information services, etc. It is going to take much longer to explore the potential of new applications and new services.
This sorting out of what is to come is one of the major functions of the initial ITV trials: to provide a platform to help us start the process of learning what interactive services people really want. We expect to see appreciably more — and more inventive — interactive content and services at the various test sites than we saw at the Western show.
A harbinger of things to come. As a way to understand what delights — or nightmares — future interactive programming could offer the home viewer, it is important to note the amazing amount of new analog programming that was exhibited on the show floor. In addition to the QVCs, ShowTimes, and E’s of the cable world, there was also The Military Channel, FAD TV, the Adam & Eve Channel, America’s Disability Channel, the TV Food Network, Romance Classics Channel and the History Channel. While these particular topics might not be what we personally hope for when we imagine a future of personalized TV, these niche-oriented traditional services are definitely a harbinger of things to come.
THE GREAT RAPPROCHEMENT
For all of this technology, the biggest topic of conversation on the show floor was not fiber and coax or interactive services, but the likely merger of the phone and cable industries. Clearly, the most dramatic turn of this year has been the sudden rapprochement between cable and phone companies.
Richard Green of CableLabs forecast this turn of events at this year’s Digital World conference when he called for collaboration between phone companies and cable companies. Still, it caught almost everyone by surprise when John Malone, the man who had carried the flag for the cable industry against the phone companies, agreed to sell TCI to Bell Atlantic.
BELL ATLANTIC’S SMITH PRESENTS AT SHOW
Not surprisingly, Ray Smith, president of Bell Atlantic, and co-architect of the Bell Atlantic/TCI merger, agrees with Green. In his remarks to the press corps during a luncheon on the second day of the show, Smith outlined how the telephone and cable companies are perfectly complementary. Cable and telephony, he said, are both mature industries, that have “reached the limits of their current business.” At the same time, both see the opportunities facing them if they could “re-invent” themselves.
Those cable companies, and those telephone companies, that do not evolve toward interactive and transactional-based services and become providers of a broad range of programming will soon be out of business, he warned.
The combined strengths of the two industries could create companies that provide excellent service (like the phone company); customer choices (like cable) and customer loyalty (which goes hand-in-hand with excellent service and low prices).
Of course the flip side — which Smith did not dwell on — was the possibility that the new cable/telco entities could continue the negative practices of each industry. Imagine a communications system with poor customer service, frequent system outages, inaccurate billing, limited programming options, high prices and poor customer service.
As far as Smith is concerned, the biggest obstacle to creating telecommunications Nirvana is ill-conceived and/or outmoded government regulation. He derided the current cable re-regulation legislation, calling it “no way to run a country’s information policy,” and called instead for a policy that was based on “symmetry and simplicity.”
The new Bell Atlantic has extremely ambitious plans: 150,000 subscribers with video-on-demand capabilities by the end of 1994, and 1.5 million by the end of 1995. “We’re going to offer cable, wireless, telephony, flossing —anything!” Mr. Smith’s dream is to have us spend most of our waking hours tied into his network.
THE MILLENNIUM JUST AROUND THE CORNER
So, where does all of this leave us? From what we have seen to date, it is easy to believe that Time-Warner, Viacom and others will be able to bring up their test sites in 1994. We should expect that there will be all manner of problems — with the networks, the content, the settop boxes, etc. That is the nature of tests — especially tests of systems as large and complex as even these modest initial installations will be. However, it looks as if the base technology will be there, and that it will work.
We have a harder time believing that this technology will be ready to roll out on a larger scale in 1995 or even 1996, as Ray Smith, his partner John Malone and others predicted. This allows almost no time to evaluate and incorporate the lessons learned from the 1994 trials. Nor does it give us any time to sort out the areas (such as user terminals) that are still in chaos, or time for settop interface standards and protocols to begin to emerge.
More importantly, it gives us precious little time to develop compelling content that will really help to drive the technology.
Everything on view at the Western show made us more convinced than ever that this phenomenon called digital media really is going to happen. It just is not likely to happen as quickly — or as smoothly — as Malone and Smith tell us.
Jonathan Seybold, Janice Maloney