F E A T U R E S    Issue 2.11 - November 1996

The Embedded Internet

By David Kline

The next big thing to hit the Net will be millions of tiny computers wired together and embedded into everything we use. This is no future vision: products are hitting the market now.



If you want to glimpse the future of the Internet (and of computing itself), look to the history of the electric motor.A hundred years ago, electric motors were relatively large "stand-alone" devices. They had to be constructed from machined parts or purchased and installed by trained mechanics. More than just a little skill was required to operate these devices.

Over the course of several decades, however, the design and operation of electric motors became more standardised, their power requirements and internal mechanisms miniaturised and their manufacture and sale commoditised. By 1918, the Sears, Roebuck and Co. catalogue offered a 5-pound Home Motor, suitable for a variety of applications, for only US$8.75 - equivalent to about $85 or £55 today.

"The many attachments shown on this page," the catalogue advertisement promised, "may be operated by this motor and help to lighten the burden of the home." And, indeed, surrounding the ad for the Home Motor were companion ads for what might loosely be called "plug-ins" or "helper applications". There was a Beater Attachment ("whips cream and beats eggs when used in connection with the Home Motor"), a Fan Attachment ("includes fan and guard, which can be quickly attached to Home Motor"), a Churn and Mixer Attachment ("for which you will find many uses"), Buffer and Grinder Attachments ("will be found very useful in many ways around the home") and, last but not least, a Vibrator Attachment ("includes three applicators and handle"). To what uses the Vibrator Attachment was put was not explained - and readers will probably find their still-living grandmothers or great-grandmothers somewhat less than forthcoming on the subject.

In those days, the new Home Motor was the subject of excited discussion at work, in cafes and around the dinner table, much as home PCs and the Internet are today. It is easy to imagine industrial age early adopters evangelising about the benefits of home motor technology to their friends and relatives ("You can beat 20 eggs in the time it takes you to do two by hand"). Curmudgeons of the day can be pictured rolling their eyes at all those frothy motorheads ("Pardon me, but who needs 20 eggs?"), just as comedians poke fun at Netheads today.

Today, of course, the Home Motor is no longer a cultural icon. Indeed, electric motor technology has become invisible, embedded inside thousands of everyday products, from hair dryers and pencil sharpeners to dishwashers and toys. Hardly obsolete, the Home Motor is instead a victim of its own success, ignored precisely because of its ubiquity. It has become a central - albeit invisible - fact of daily life.

Can it be that, as the electric motor went, so goes the Internet? More to the point: what will it take for the Net to become so embedded in social and economic life, so central to everyday communication and commerce, that it becomes as invisible and ubiquitous a feature of daily life as the electric motor?

We may have the answer to that question sooner than we think. A little-known technology, far more significant than Java, is about to be deployed on the Internet. This technology could rapidly transform the Net into a medium with genuine mass-market penetration in our society. It will not only greatly expand the Net's role as a mass consumer medium, but will also change it into a powerful industrial force which, by the end of this decade, will reshape the market dynamics and the competitive strategies of businesses.

This miracle technology is called "embedded systems": tiny crash-proof computers that are embedded or hardwired into everyday products and dedicated to the performance of specific tasks or groups of tasks. Already used in a host of industrial and consumer products, from anti-lock brakes to VCRs to microwaves, embedded systems typically offer significantly faster, cheaper and much more reliable real-time performance than the cumbersome multipurpose or "fat" software used in PCs. In fact, 90% of the world's microprocessors are used not in PCs, but hidden inside common household or electronics products.

Now, thanks to new technical advances made by embedded systems developers, these invisible computers have become Internet-ready. Embedded software now offers automated Net connectivity with about one-hundredth the memory that Windows and other PC operating systems require, at about one-tenth the cost.

For David St. Charles, president and CEO of Integrated Systems Inc. (ISI), the $100-million market leader among embedded systems developers, this sort of cheap Net connectivity opens up enormous new possibilities. He offers a vision of a truly universal Internet, in which the common artifacts of daily life - a car, a TV, a CD player, a phone, a photocopier, a natural gas meter, a PC - are all connected via cheap automated software in a global network he calls the "embedded Internet".

"This is the next stage," St. Charles says in his measured, Canadian voice. "This is where we make the Internet real. And I mean as ubiquitous as electric motors or telephones, where all sorts of devices and systems are linked invisibly together. Where people and devices easily and automatically communicate with each other with no one having to know anything about computers or software or TCP/IP stacks or anything else. It is everywhere, it does everything and it's absolutely a no-brainer to use. The push-button Internet!"

Here, at last, is a vision of the Internet for the masses, in their hundreds of millions and ultimately in their billions. In it, the Net is no longer just a publishing or an entertainment or a personal communications medium, but rather a fundamental and indispensable engine that drives social and economic life. It is an industrial medium that enables automated monitoring and reporting on factory-floor production; a home security and emergency response medium far more reliable than today's phone-based system; a medical medium through which patient treatment plans are automatically routed to relevant providers; a consumer appliance and office equipment medium that checks the status of devices and initiates electronic repairs; a utility management medium in which power usage is read and managed remotely. You name the application, the Net will be essential to it.

In this vision is, at last, an Internet set free from its PC-centric straitjacket; a cyberspace transformed from just another platform into an omnipresent glue that binds the whole of society, with all its trillions of daily social and economic interactions, into a truly connected civilisation.

But the secret to all this, insists St. Charles, is invisibility. "If you want the Internet to be everywhere," he says, "it has to be visible nowhere. It has to be unseen, unnoticed, undiscussed."

Disappearing act

"Unseen" is certainly not a word we would normally use to describe either the Net or the computing devices we use to access it. As Interval Research Corp.'s Brenda Laurel so adroitly puts it, using computers is like going to the cinema and having to watch the projector instead of the movie.

In similar fashion, St. Charles regards the current iconisation of the Internet in media and cultural circles as a sign, not of its assimilation into social and economic life, but of how far it still must go to become a medium of commerce and communications for the masses. In his view, we will know that the Net has truly arrived when there are no more Internet cafes, no more Internet seminars and no more Internet magazines (or, for that matter, articles such as the one you are reading now).

This makes sense when you look at technologies that have achieved genuine mass-market penetration in our society. After all, you don't see any magazines called Dial Tone Today featuring the latest ins and outs of the Net's closest cousin, telephony. In fact, we hardly even see the telephone anymore - we simply use it, thinking only of what we want to do with it. The technology, in direct proportion to its becoming a central fact of daily existence, has become invisible.

"Invisibility, you might say, is my business," says St. Charles. "It's what I do for a living - I make technology disappear."

He is not kidding. Once called the "Microsoft of hidden computers" by Forbes magazine, ISI's technology appears to be everywhere. It is embedded inside Sony direct broadcast satellites, Motorola communications satellites, Kodak and Xerox printers and copiers, Hewlett-Packard interactive TV set-top boxes, Sega karaoke systems, Gilbarco automated gas pumps, Boeing planes, Avis rental car navigation systems, Philips consumer appliances and AT&T, Northern Telecom and Alcatel switching equipment, to name but a few.

"Most people have no idea what our industry does or how deeply our work affects their lives," St. Charles notes, "despite the fact that in many ways it has probably done more than Microsoft and the whole PC industry to bring computing technology into people's day-to-day lives. If you ask people how many computers they use, they'll probably say one - their desktop PC. But the reality is, it's probably more like 15. They use computers to drive the tape in their VCRs and they use them in their microwaves, cars, automatic teller machines, gas pumps and phone systems. You name it, these days it's probably got an embedded computer in it."

So what exactly is an embedded computer? Very simply, an embedded computer is a microprocessor, a real-time operating system and application that has been built into a larger product to handle the control functions of that product. Typically, an embedded systems company delivers just the master software - the real-time operating system and application code - to the end-product manufacturer, which buys and installs the micro-processors and other hardware components.

The advantage of embedded systems is that, partly because they are dedicated to the performance of a single task or specific group of tasks, they can be optimised to deliver a level of speed, reliability and cost that PC hardware and software manufacturers could never in their wildest dreams hope to achieve. And as for ease of use - let's just say that the PC industry is not even on the same planet as ordinary consumers.

Only in the PC industry is it considered customary to sell products so unreliable and confusing that nearly a third of buyers are unable to use them without help. According to Dataquest, some 200 million calls to PC technical support hotlines will be made this year. And of the calls made during peak hours, only one in seven will even get past a busy signal. In contrast, when was the last time you had to "install" your TV, your phone, or your VCR? And when was the last time you had to call technical support because your microwave wouldn't work or you couldn't figure out how to use it?

Truth be told, the PC is radically different in design from any other technology or appliance used by the general consuming public. Indeed, if Intel and Microsoft had designed our kitchens, we'd probably all be using $3,000 multipurpose "Kitchen Processors" rather than the low-cost, dedicated, push-button appliances we now have. These Kitchen Processors, of course, would need to be configured and launched to, say, heat coils for a toasting application, and then reconfigured and relaunched to heat incandescent filament for a lighting application.

As the basis for a truly ubiquitous Internet, today's PC simply will not cut it. But will the embedded Internet approach be realised only in the far-distant future?

Apparently not, for Integrated Systems has already built Java and HTML capability into its open embedded operating system, known as pSOS. And the company has already signed a deal with Philips for an under-$300 Net-communicating browser appliance. It is also in talks with a number of other telecom, industrial and consumer electronics firms to embed Net-ready computer systems into everyday consumer products, from smart kitchen appliances that communicate among themselves to automated automobile diagnostic systems that flag mechanics about required repairs.

And ISI is not the only company bringing its hidden computer expertise to the world of the Net. Iowa-based Microware Systems Corp. has also announced an embedded Internet strategy and is working with Mitsubishi on a TV with Web capability integrated into its circuitry for release this autumn. This Web TV enables VGA-quality graphics, something that cannot be done via a Web box atop the TV, and it will cost only $200 more to produce than regular TV models. Microware has also signed with Uniden to produce an Internet-capable phone that collects email, and with Hongkong Telecom IMS to deliver a Java-enabled version of DAVID, a highly regarded digital TV platform, to provide online services over television to homes in Hong Kong.

The Industrial Net

Even more significant than the consumer uses of embedded Internet technology may be the industrial potential. Indeed, systems that use the Net to reduce industrial costs and streamline commercial operations (the "Industrial Net", if you like) are likely to deliver far, far more in the way of real bankable revenues in the next few years than will consumer applications.

ISI is working with Xionics and other office equipment firms to embed low-cost Net capability into millions of printers and other office peripherals. These firms hope to achieve multimillion-dollar savings in labour and service costs by performing remote equipment diagnostics and maintenance via the Net.

How would each printer get connected to the Net? It could be through phone lines, electrical wiring, a LAN or even a "spread-spectrum" system using low-power radio signals to feed the data to a wired or wireless gateway.

And why use the Net rather than simple phone lines? Only the Internet can serve as the infrastructure for a truly connected world. As a medium, it is much cheaper than the phone network. It provides standardised interoperability between all types of disparate devices, networks and systems. The Net is also a better platform than the phone system for adding intelligence and agent technology. And the Net's distributed architecture makes it more efficient at rerouting communications around damage and bottlenecks.

But perhaps most importantly, the Net's packet-switched approach to handling communications is simply better suited to handling the billions or even trillions of relatively short data transmissions likely in a future connected world than the phone network is, where each connection essentially requires a dedicated link for the entire duration of the transmission.

One can easily imagine a host of other industrial possibilities. Remote meter reading over the Net by power utilities? Net-communicating smart sensors to remotely manage factory production processes? Low-cost Net-connected alarm systems for the fast-growing home security industry? The potential of embedded Internet technology for industrial uses is enormous.

But beyond its many industrial and consumer applications, a truly Net-connected world requires that the medium be woven into the basic infrastructure of society, including our public services.

"Think about 911," suggests Microware's founder and CEO, Ken Kaplan. "Everyone has access to it and it's pretty much basic to societal functioning, right? Well, I can foresee a really cheap 911 device that costs just a few dollars and has only one button on it - an embedded Internet pager-like gadget with a GPS (Global Positioning System) chip and a wireless connection to the Net. You get mugged or have a heart attack - just push the button and help is on the way. It could even be self-actuating if you use a smart sensor to monitor heart activity. The technology already exists for it. All we need is cheap connectivity through a ubiquitous, low-cost, public switched-network that can be automated and doesn't require voice or other human intervention - in other words, the Net."

"You know, maybe I shouldn't have mentioned this idea of a 911 device," worries Kaplan. "I mean, this could be the killer app of the embedded Internet and here I've gone and given it away."

However, he needn't worry about competition, at least from the ranks of PC hardware and software vendors. To build a 911 device linked to universal Internet will require design characteristics and product features that are simply not intrinsic to the PC industry. Chief among these are low cost, fast performance, utter simplicity of use and total crash-proof reliability.

Unlike multipurpose PC operating systems such as Windows, which must take into account every conceivable user input and be able to support a huge variety of applications, an embedded system typically must support only a narrow suite of functions and needs to respond to only a very limited range of human input - like a foot depressing a car brake pedal to activate a computer-controlled anti-lock braking mechanism. As a result, embedded systems have far smaller memory requirements. ISI's Internet-ready version of its pSOS operating system uses 16Kb of memory, which is about one-hundredth the memory used by Windows 95.

This enables embedded systems to offer a far more economical approach to building cheap, Net-browsing appliances than those touted by traditional PC industry vendors, such as Oracle with its stripped-down network computer. A full embedded system - hardware and software - that uses Microware's OS-9 operating system only costs about $150. That includes a Net-ready operating system (at $10 a pop), a 32-bit PowerPC processor, and all the components needed for video, input and output ports, a graphical user interface and a browser. Not bad.

For products not requiring a user interface or input-output extensions, the cost can be even less. An Internet-ready system embedded in your gas meter that enables the power company to conduct remote meter-reading could cost just a few dollars, as presumably would Kaplan's 911 device.

Wiring Joe Six-pack

The superior performance of embedded systems also derives from the fact that their operating systems are real-time operating systems, meaning that Microware's OS-9 and ISI's pSOS react in a rapid and consistent way to events that occur outside the computer, with little if any delay imposed by the computer itself. This is partly due to the entirely different design methodology that embedded computer companies have adopted to sell systems to companies that use them in mission-critical applications.

Consider, for example, that when a PC user types a keyboard command and the system hangs for a moment, it is at most an annoyance. But when a driver speeding along a rain-slicked freeway at 65mph slams his foot on the brakes and a wheel locks up, any delay by the computer in modulating the brakes could be a fatality.

"Bugs?" says Microware's Kaplan. "For us they're a disaster. For PC software companies, they are an additional revenue opportunity." Meaning, of course, that this year's software bug is next year's purchasable upgrade.

Indeed, Kaplan argues, few people realise how different embedded software design is from the approach used in the desktop PC industry. "In the PC world, software design is an iterative process. Back and forth it goes, from the programmers to the internal test group and then back to the code writers, until it's more or less ready to test on beta users. Microsoft sent its pre-release Windows 95 OS out to about half a million beta testers. Those people managed to catch some of the bugs but, as we all know, not all. Well, there's no way in hell that we can send beta software to Boeing and then wait for planes to start falling from the sky to discover the bugs."

According to Kaplan, each stage of the embedded systems design process has specification and quality objectives that absolutely must be met before the code moves on to the next stage - an approach more typical of aerospace firms than PC companies.

Translate those very different design methodologies of the PC and embedded systems industries to the world of Net-enabled 911 devices imagined by Kaplan. Would you trust your life to an emergency-response device built by a PC vendor, even assuming it could be built at a consumer price point? There does not seem to be much point in having a 911 device if it is going to crash from time to time.

The need for mission-critical design is not limited just to applications where lives are at stake. "Even in the world of everyday consumer electronics," notes Kaplan, "customers simply will not put up with the glitches and breakdowns and user confusion common to PCs. Anyone who wants to bring the Net to the masses had better understand that."

Furthermore, when we imagine Net-enabled applications used by hundreds of millions of people, there must be a degree of social assurance - much like the certainty of hearing a dial tone when we pick up the phone - that somehow, in the background somewhere, the whole underlying system is doing its job and performing as promised.

The invisible Internet does not mean the end of PCs - indeed, the ranks of today's fully featured PC users will continue to grow. Nor does it mean that Net capabilities must be dumbed down to only those applications and uses possible in a simple push-button device. The concept of the embedded Internet simply recognises that epochal technological change generally comes to the world on the world's terms, shaped by human nature and our existing political economy. Slowly and imperceptibly at first, new technologies also reshape us. But the starting point for tomorrow's great technology-induced social changes must be the masses of technologically unsophisticated ordinary consumers.

"These are the people who need to become wired if the Net is really to grow into the enormous social and economic force it is capable of becoming," insists Kaplan. "You say the Net will change the world? Well, those people are the world. So you had better design a Net-connected world that ordinary people can really understand and use." In other words, instead of trying to get everyone to buy a PC and master its arcane methodology, simply put Net connectivity inside the everyday devices and systems that people already use. Bring the mountain to Mohammad, as it were. ISI's St. Charles: "Embedded systems are the only way to make the Internet as cheap, easy and transparent as the electric motor or the phone. It's the only way to make the Net a basic and ubiquitous fact of life. It's that simple."

David Kline is a contributing writer to Wired, commentator for National Public Radio and co-author of Road Warriors: Dreams and Nightmares Along the Information Highway.