| Technically impressive, Apple's Newton MessagePad and the Tandy/Casio Zoomer are the first true Personal Digital Assistants. Their potential is tantalizing, but the reality leaves something to be desired. Prices are high, handwriting recognition is marginal, and communications are incomplete. Yet with technology advances, PDAs may yet fulfill their original vision. |
| The first personal computers all looked the same: a box
with a video screen and a keyboard. Beneath the surface,
however, were significant differences in terms of
microprocessors, operating systems, and system
architectures. Likewise, the first pen-based PDAs
(Personal Digital Assistants) all look very much the same,
too: a tiny box with an LCD screen and a stylus. But appearances, once again, are deceiving. Some devices that claim to be PDAs aren't PDAs at all without absurdly stretching the definition. They're aimed at widely different markets, and they have little in common but portability. Their differences become even more apparent after a technical analysis of their underlying architectures. Some PDA-type devices are scaled-down desktop computers, while others explore bold new territory. Constraints of cost, compatibility, memory, CPU power, size, weight, and battery life have forced their designers to make very different sets of compromises and trade-offs. Even if you isolate the two devices that appear almost identical — the Apple Newton MessagePad and the Tandy/Casio Zoomer — you discover that their supporting architectures are sharply divergent, though perhaps in ways that won't become obvious to casual users until future iterations. At the same time, the early PDAs also have several things in common. They're all rather pricey compared to other electronic gadgets aimed at the mass consumer market. They all squeeze large amounts of computing power into amazingly small packages, but in some cases it's not enough power to satisfy their ambitious aspirations (handwriting recognition in particular). They all come with a suite of useful applications, such as notepads, address books, and appointment calendars, but no "killer application," such as VisiCalc or WordStar, that dramatically reveals new possibilities. Are PDAs ready for prime time? Are they worth the price? Can they justify their existence by doing anything important? Or will they be doomed to early failure, banished to their design labs until the technology catches up with our expectations? This partially depends on what you expect from a PDA and how much you're willing to pay. Most people will probably be disappointed with the early PDAs or will guard their wallets until prices drop. But a remarkable number of companies are gambling on hopes that thousands of people will think PDAs are the best thing since sliced silicon. PDAs are seeking a niche, but they're not just niche products. They're ultimately aimed at the millions of people who will probably never buy a personal computer, and they're destined to have a major impact on personal computing at large. PDAs already are accelerating the trend toward wireless communications and will encourage the on-line distribution of software. More important, one of these tiny devices adopts a new approach to system software that foreshadows the object-oriented operating systems that will be coming to desktops later in the decade. What's a PDA? Not surprisingly, there is no industrywide consensus on what constitutes a PDA. Because Apple chairman John Sculley coined the term in January 1992 and guided the Newton architecture's development, it's tempting to let Apple define the grounds of this debate. However, that wouldn't be fair to Apple's competitors, and it may not take into account the expectations of users. Apple's MessagePad, the first in a family of products based on the Newton architecture, is the most interesting early example of the genre. If successful, the MessagePad will be widely imitated because, unlike the proprietary Macintosh, the Newton architecture is openly licensed to all comers. The list includes Cirrus Logic, Matsushita, Motorola, Sharp, and Siemens. Sharp, which manufactures the MessagePad for Apple, is also selling a variant called the ExpertPad. However, Sharp is hedging its bets with a tablet-size device called the PT 9000 that isn't based on Newton technology (see the text box "Sharp's Non-Newtonian PDA"). Instead, the PT 9000 uses the GEOS operating system from GeoWorks (Berkeley, CA). Another important GEOS-based system is the result of a partnership between Tandy (Fort Worth, TX) and Japan-based Casio, with significant input from GeoWorks and Palm Computing (Los Altos, CA). Nicknamed the Zoomer (the Tandy version is actually called the Z-PDA; the Casio model is called the XL-7000), it is perhaps the MessagePad's closest competitor in terms of appearance, user interface, and intended audience (see "Ease of Use Is Relative" on page 89). Eo, Inc. (Mountain View, CA), makes the Eo 440 and 880 Personal Communicators, a pair of clipboard-size pen computers based on AT&T's Hobbit chip and Go Corp.'s PenPoint operating system. The Eo 440 and 880 perform many of the same functions as the MessagePad and the Zoomer, and in some ways they do a better job of it. For instance, their larger screens are more suited for creating documents, they can receive faxes as well as send them, and they can be equipped with cellular phones. Nevertheless, the Eo devices are not PDAs. Unlike the Zoomer and the MessagePad, they make no pretense about being consumer products. Other contenders in this category include the Hewlett-Packard 100LX, a DOS-based palmtop computer; pocket-size personal organizers, such as the Sharp Wizard OZ-9600 and the Casio Boss; and content-oriented reference tools, such as Franklin's Digital Books. For various reasons, however, they fall short of what we expect from a true PDA (see the table "What Makes a PDA?" on page 72). Sculley risked inflating those expectations over the past two years by making Newton the most visible work-in-progress in Apple's 17-year history. But users' expectations have also been influenced by Gene Roddenberry, the late creator of Star Trek. Roddenberry first introduced PDAs to the public in 1966. When Star Trek was still in its conceptual stages, he decreed that no paper or pencils were to appear anywhere on the sets of the starship Enterprise. Instead, crew members jotted notes on electronic tablets that eerily resemble an Eo 880, and they used a hand-held device called a tricorder to gather, process, and display information. On Star Trek: The Next Generation, the sequel to the original series, crew members communicate remotely by speaking into active-badge-like devices worn on their tunics. This futuristic vision of multifunction hand-held computers, wireless communications, and an effortless user interface is the inspiration for today's PDAs, even if some designers do not acknowledge or are not consciously aware of the pop-culture source. Unfortunately, they have to meet twenty-fourth-century expectations with twentieth-century technology. Still, the vision yields a practical definition of a PDA. At a minimum, a PDA should have significant computing power and be capable of performing a number of different functions, which rules out high-end calculators and dedicated language translators. It should be easily held in one hand, usable almost anywhere while you are standing or sitting, and capable of running on battery power for extended periods of time, which rules out laptop and notebook computers. It should offer flexible communications, at least within the bounds of today's rapidly evolving but still-limited infrastructure, which rules out most personal organizers. And, most challenging of all, it must be so affordable and easy to use that it will attract the millions of people who have thus far steadfastly resisted desktop PCs. "We're not talking about a downsized personal computer," says Michael Tchao, manager of product planning and strategy for Apple's Newton Group. "We're not talking about Lotus 1-2-3 in the palm of your hand. We're not talking about pocket PageMaker. We're not talking about a little-bitty Microsoft Word with little-bitty keys and little-bitty menus. We're talking about a new class of device and a new category of users." However, no device currently available meets all these criteria, especially at an affordable price. Like color TVs, VCRs, CD players, and other consumer electronics gadgets, PDAs won't become genuine mass-market items until prices drop below $300, and ideally even lower. The Mighty Pen Not everyone will agree with this definition, however, particularly those whose PDA-like products it excludes. For example, the Eo 440 and 880 are capable machines, but at $2000 to $3300, they're clearly not intended for mass consumption. The HP 100LX and Sharp Wizard are often considered PDAs, but they have QWERTY keyboards. Although there's disagreement on this point, those who target the mass market convincingly argue that keyboards are simply too intimidating for the computer-illiterate. "If you're designing a device for broad consumer acceptance, you've got to have a pen interface," says Joe Ratner, Zoomer project manager for Radio Shack. "From the very beginning, we decided the Zoomer had to have a pen interface. It couldn't have a keyboard." Ratner explains that it's impossible to fit a touch-typeable QWERTY keyboard on a pocket-size device, and crowded keyboards that have dozens of multifunction keys are even more intimidating to consumers than a full-size desktop keyboard. His counterparts at Apple, who interviewed more than 1200 consumers while designing the MessagePad, came to the same conclusion. Gerry Purdy, vice president and chief analyst for mobile computing at Dataquest (San Jose, CA), predicts that QWERTY-based devices such as the 100LX and Wizard are "going the way of the slide rule:" PDAs will eventually replace them. The point is that PDAs aren't substitutes for PCs; they're different devices aimed at a different market. "I'm not expecting to replace my desktop computer or my laptop with a PDA, but I'm definitely going to have a PDA," says Jeff Hawkins, founder of Palm Computing, who first conceived the Zoomer while working at Grid Systems. "For heavy-duty text writing, pen input won't replace a keyboard." PDA designers think their devices are more likely to be used by consumers of information than by creators of information. Therefore, user input will tend to be brief and frequently limited to selections from menus. Those who need to enter lots of data will use conventional computers, although some PDAs, such as the Sharp PT 9000, will offer keyboards as an option. "But a pen interface doesn't necessarily mean the same thing as handwriting recognition," Ratner points out. "For some applications, you don't need recognition. Digital ink is good enough." Digital ink — storing the image of the user's handwriting instead of translating it into ASCII text — is one of the trade-offs juggled by designers of today's PDAs (see the text box "Ink vs. ASCII"). PDAs tend to look alike, so one way of telling them apart is to peel back the surface. Their underlying architectures determine not only what they can do now, but also what they'll be capable of doing in the future as the technology advances. Operating Systems of the Future Some operating systems running on PDA-type devices are descended from those of desktops: DOS, GEOS, Windows for Pen Computing. Others are dedicated to proprietary hardware and don't attempt to duplicate the functionality of a full-fledged operating system: These include those of the Sharp Wizard and Casio Boss. Some operating systems, such as Magic Cap from General Magic and WinPad from Microsoft, look promising but haven't yet appeared in actual products. Only two operating systems currently available were designed from the ground up to support PDAs: Go's PenPoint and Apple's Newton Intelligence. Both are heavily object-oriented and depart from the traditional model used by today's operating systems for desktop computers. Newton Intelligence borrows a few of its components from the Mac: Portions of QuickDraw (the Mac imaging engine) handle the MessagePad's display, and an AppleTalk protocol stack supports name lookup, zones, data streams, and printing, although there's currently no support for AppleTalk Remote Access. Newton Intelligence is multitasking and extensible, and Apple plans to distribute software updates and extensions via on-line services. The Newton operating system consists of four major components: the Recognition Architecture, the Communications Architecture, the Information Architecture, and the Intelligent Assistance. Most of it is written in C and C++, with a small portion of the low-level kernel written in assembly language. The user-interface layer is coded in C, C++, and NewtonScript, a new language that combines features of C and Pascal. NewtonScript is the development tool for all Newton applications software, including the programs that are burned into the MessagePad's ROMs. Drivers and other low-level functions that access the MessagePad's hardware directly are best written in assembly language. The Recognition Architecture has a recognizer engine that translates your printing, cursive writing, or any combination of the two. The recognizer was designed by ParaGraph International, a U.S./Russian joint venture based in Moscow and Sunnyvale, California (see "Pen Computing Catches On" on page 105). It's trainable, automatically adapting to your handwriting style over time. Similarly, a graphics interpreter detects specific objects and penstrokes as you sketch on the display. For example, it automatically transforms a round scribble into a perfectly formed circle; four perpendicular lines become a precise rectangle. Both the handwriting recognizer and the graphics interpreter detect several predefined gestures for manipulating words and shapes. You can move objects, resize them, and cut and paste them on a system clipboard. Newton's Communications Architecture supports a wide variety of I/O devices via the MessagePad's RS-422 serial port, PCMCIA slot, and 19.2-Kbps infrared transceiver. The serial port connects to printers, fax modems, and desktop PCs. The Type 2 PCMCIA slot accepts wireless communication cards and other add-ons. The infrared link lets you beam data to another MessagePad or a Sharp Wizard OZ-9600. This architecture is extensible, so new communications options can be added in the future. A Central Data Bank What sets Newton Intelligence apart from the typical desktop operating system is the Information Architecture. It's datacentric, not filecentric, a radical departure from DOS, System 7, Unix, and other common operating systems. PenPoint has a very similar structure, although it doesn't go quite as far as Newton in abandoning the old file-oriented model. Both Newton and PenPoint give us a likely preview of the object-oriented operating systems from Microsoft (Cairo) and Taligent (Pink) that are coming to desktops. Instead of storing data in discrete files organized in a hierarchical file system, Newton Intelligence lumps everything together in an object-oriented central repository that's accessible to all applications. Technically, this is known as a persistent object store, and it promotes data-sharing among applications without the complex plumbing that's now being retrofitted to operating systems for desktops. Because conventional operating systems segregate information in files, applications can't share live data until the operating system is equipped with a mechanism such as Microsoft's DDE or OLE, or Apple's Publish/Subscribe or Amber/OpenDoc. Although these mechanisms work, they add yet another thick layer of code that wasn't anticipated in the original architecture of the operating system. They also force developers who want to take advantage of the new features to modify their code to expose internal objects to other applications. If the code isn't already object-oriented, extensive rewriting may be necessary. Both Microsoft and Apple are in the process of pulling their respective developer communities in these new directions. Newton's Information Architecture takes a different approach to data storage and sharing. When data enters the MessagePad, the Information Architecture automatically tags (i.e., names), compresses, and saves the data in an object called a frame. A frame is a structure composed of tagged locations called slots. Slots can contain user data, program code, and even other frames. As objects, high-level meanings can be attached to frames: for example, address can represent "1 Phoenix Mill Lane." A collection of related data frames is known as a soup, and soups reside in physical locations called stores. For instance, a soup of contact names might exist in a store located on a PCMCIA RAM card. Soups automatically maintain indexes to their frames. This unified model means that all information stored in the MessagePad automatically becomes part of an object database that Newton applications can search, modify, and display in countless ways. It eliminates the overhead of data translation and the redundant copies that often result from translation. Furthermore, because all information is processed as Unicode, the MessagePad readily adapts to foreign languages. It's in the Soup All of this dovetails with the final component of Newton Intelligence, the Intelligent Assistance. This component takes advantage of the data soup to make plausible connections between the information and your actions. For example, if you ask the MessagePad to "fax to Lisa," the Intelligent Assistance automatically assumes the sketch you just drew on the screen is what you want to send. Next, it locates all people named Lisa in the object database. If there's more than one, the MessagePad prompts you to make a choice. Then it retrieves Lisa's fax number and generates a fax document. The next time you connect your modem to a phone line, the MessagePad sends the fax. The Intelligent Assistance is also extensible. Although all these capabilities can (and will) be bolted onto existing operating systems, Newton Intelligence was designed around this datacentric model in the first place, making it considerably more compact and efficient. For example, it easily fits within the MessagePad's 4 MB of ROM, leaving ample room for several other system components and all the built-in applications software. Different Approaches to Objects PenPoint is built on a similar object-oriented structure, but it doesn't appear to go as far as Newton toward discarding the traditional file-oriented model. PenPoint applications can store data as objects, much like Newton does, but they also have the option of saving documents in conventional file formats such as WKS or RTF. A systemwide import/export architecture translates those files as necessary when they are called by a PenPoint application. Newton applications, on the other hand, might provide backward compatibility with conventional file formats by encapsulating a file within a slot or a frame in the object store. All these structural details will probably remain transparent to users — unless something goes horribly wrong. That's when the subtle differences between Newton and PenPoint might become more obvious. Newton critics say that its centralized object database could backfire by compromising data integrity and that PenPoint has better safeguards in this respect. Data integrity is paramount in a RAM-based system, because a crash could corrupt all of a user's accumulated data, not just a single file associated with a single application. PenPoint was designed with extensive memory protection, so applications are effectively "walled off" from each other. The walls keep one application from interfering with another application's data or from crashing the whole system. To keep those thick walls from impeding IAC (Interapplication Communication), PenPoint has a system-level messaging mechanism that allows PenPoint applications and their objects to interact with each other. Memory protection is receiving close attention in the design of multitasking operating systems, and it's a key feature of Microsoft Windows NT and IBM's OS/2 2.1. But Apple says the kind of memory protection found in PenPoint adds too much overhead to data-sharing between applications and that Newton's Storage Manager protects the integrity of the central repository in better ways. Newton applications never access the object database directly, explains Larry Kenyon, principal engineer of Newton Intelligence. Instead, all reads and writes are handled as transactions and funneled through the Storage Manager's I/O system, where a sentry verifies the data before updating the object store. "Applications can't even touch the data," says Kenyon. "The sentry is always in place to keep that from happening." Even if a system crashes or is switched off during a transaction, no data will be corrupted, he claims. Of course, no operating system can prevent loss of data if the hardware is physically damaged or if the MessagePad's main batteries and lithium cell go completely dead. Prudent users still need to back up their data. The NewtonScript development language provides additional safeguards, says Apple's Michael Tchao. Unlike C, it automatically handles memory management and garbage collection. Applications programmers no longer have to allocate memory, and all references are to objects, not to handles or pointers. Tchao says 80 percent of crashes are caused by memory management problems and that NewtonScript makes the programmer's nightmare of memory leaks and dangling pointers a thing of the past. NewtonScript objects also have latent typing, and all operations are type-checked before they're performed. This nails argument errors before they have a chance to trash the stack and cause problems dozens of instructions later. Even if NewtonScript makes life easier for programmers and contributes to data integrity, will developers embrace the new language? Critics say that because NewtonScript is nonstandard and bucks the industry trend toward C and C++, porting code won't be as easy. But Tchao predicts that programmers will adapt quickly to NewtonScript's hybrid C/Pascal syntax. Based on early experience, he claims development will be faster on the Newton than on other platforms, mainly because there's less code to write. Developers won't have to recompile their applications to run on Newton devices built around other microprocessors, because NewtonScript code is executed by a run-time interpreter. Apple has already received inquiries from about 1500 "serious" developers, says Tchao, and about 50 applications written in NewtonScript will be introduced by Christmas. By the end of 1994, Apple predicts that about 300 applications will be available. "Anytime you invent a new language, the language zealots — and there are a lot of language zealots out there — run screaming from the room," says Tchao. "But we haven't found that to be the case with NewtonScript." A New User Interface Another consequence of Newton's datacentric architecture is that it discards the old data-processing model of booting an operating system, running and quitting application programs, loading and saving documents, and maneuvering through hierarchical file systems that mirror the layout of mass-storage devices. Although this model has served us well for decades, it assumes a certain familiarity with a computer system's underlying structure — the difference between an executable file and a data file, for instance, and the hierarchy of directories, subdirectories, and files. If PDAs are to successfully penetrate the mass market, these assumptions have to be abandoned. The MessagePad has no start-up sequence, and you never explicitly load or run a program on it as you do on a desktop PC. Instead, you simply flick on the power switch, and you automatically return to the same application and document that was active during your last session. You move smoothly from one application to another by tapping its icon. To the casual user, it appears that all applications are always running, although actually the operating system is transparently swapping chunks of code in and out of the computer's RAM (640 KB on a standard MessagePad). When you move to a different application, you find yourself at the exact spot in the last document that you created with that application. To switch documents, you simply tap on an icon. You never have to explicitly load or save a document; any changes you make are automatically updated in the central database. Nor must you deal with a file system that exposes the detailed structure of the system's mass storage. You can call up a scrolling list of notes that you've created, but you can't view a "disk directory" that mixes data files with executable and auxiliary files. Newton saves documents chronologically, and you can organize them into folders if you want to do so, but the nuts-and-bolts framework of the storage system always remains hidden. This is a greatly abstracted extension of the Mac's file system, which shows only one icon for an executable file, even though it actually consists of separate files for the resource and data forks. PDA designers consider such details extremely important. Apple says it went so far as to hire a cultural anthropologist to observe people and notice how they use such things as Post-it Notes and other traditional tools. "You have to go beyond ease of use and provide assistance to the user," says Shifteh Karimi, user studies manager for Apple's Newton Group. Again, PDAs are giving us a peek at the future. Desktop computers are moving toward the same datacentric philosophy, although retrofitted solutions such as OLE 2.0 and Amber/OpenDoc are not as elegant. Desktop PCs could also benefit from Newton's abstract file system, as mass-storage devices with gigabyte capacities strain users' ability to find individual files in deeply nested stacks of folders and subfolders. The Zoomer's Architecture Apple is by no means alone in recognizing the value of a new user interface for a product aimed at the masses. The Zoomer takes a strikingly similar approach, even though it's built atop an entirely different architecture. While Apple spent millions of dollars infusing the MessagePad with the latest technology — an advanced operating system, an optimized development language, an ARM610 (Advanced RISC machine) RISC chip — Tandy tried to reduce costs by building the Zoomer with tried-and-true parts. Casio designed an Intel-compatible CPU that's equivalent to a 7.5-MHz 8088. GeoWorks adapted its desktop GEOS environment to reside in ROM. Palm Computing wrote almost all the applications software in assembly language and created an extremely compact handwriting recognizer that fits in 48 KB. Intuit contributed Pocket Quicken, a mobile version of its best-selling money management program. But despite their different evolutionary paths, the MessagePad and the Zoomer could appear very much alike to casual users. Not only do they look like twins, but they'll also hit the street at about the same price ($700), and both conceal their internal complexities with remarkably similar user-interface layers. Both the MessagePad and the Zoomer are instant-on/instant-off devices that automatically drop you at the same spot in the same document from your previous session. Both allow you to move transparently between applications by tapping on a row of "hard icons," affixed to the bottom of the screen, or "soft icons," displayed on the LCD. Both shield you from their underlying file systems, although the Zoomer lets advanced users run a File Manager that exposes the realities of DOS. Both have similar communications capabilities and a suite of applications in 4 MB of ROM. Structurally, however, they're worlds apart. GEOS has high-level objects that save a lot of coding for applications programmers, but it has nothing like Newton's object-oriented database. Applications share data the traditional way: by exchanging (and, if necessary, translating) discrete files or by sending messages. Although GEOS is not chained to the DOS file system, it maintains the link to DOS for file compatibility. Yet the Zoomer cannot run DOS programs; with the exception of Grid PenRight applications, the ability to run DOS executables was deliberately disabled because of limitations imposed by the 320- by 256-pixel screen. Naturally, GeoWorks CEO Brian Dougherty doesn't view these differences as a handicap. The Newton architecture is unproven, he points out, and GEOS provides a rich operating environment for PDAs that is being widely licensed to several vendors (including Apple's Newton partner, Sharp). Programs written for the desktop version of GEOS can be ported to the Zoomer without recompiling, because GEOS decouples the user interface from the application code. The PDA version of GEOS crams plenty into 1.5 MB of ROM: about 40 printer drivers, an imaging engine, IAC, multithreading, on-line help, PCMCIA support, high-level objects for programmers, and peer-to-peer networking via the serial and infrared ports. Approximately 1 MB of this ROM-based code must be bank-switched into RAM, because only the low-level kernel is XIP (executable in place). Dougherty says the next version of GEOS will keep more XIP code in ROM to minimize bank-switching, thus taking some load off the slow CPU. If the Zoomer's 8088-class CPU wasn't limited to real mode, GEOS could run in protected mode, mapping the ROM into flat address space. With additional RAM, this could eliminate bank-switching altogether and speed up the machine considerably. Dougherty would love to see this happen. In fact, from the very beginning he pushed Casio to build the Zoomer around a more powerful 80x86-compatible CPU, such as Chips & Technologies' PC/Chip, but he says Casio insisted on using a less powerful chip to maximize battery life. However, future versions of the Zoomer could escape that trade-off. Intel and VLSI Logic are developing a family of highly integrated chips that combine the logic cores of the 386SL and 486SL with other system components to make a two-chip solution for 80x86-compatible PDAs (see the text box "PDA CPUs: New Form Demands New Functions"). Neither Tandy nor Casio has announced plans to use the Intel/VLSI option, however. Early Adopters or Suckers? Although the MessagePad and the Zoomer fall short of some expectations and are relatively costly, from a purely technical standpoint they are both undeniably impressive. Compared to their distant ancestors that also broke new ground — the Apple I and the Tandy/Radio Shack TRS-80 Model I — they are remarkably mature products. They also have bigger ambitions. With sub-$1000 prices, soon to drop below $500, there's not much profit in PDAs. To succeed as mass-market consumer items, PDAs ultimately must move in volumes measured in the millions. Apple in particular has a lot riding on this gamble. Newton is one of Apple's answers to Microsoft's "Windows everywhere" strategy (see "Windows, Windows Everywhere?" in the June BYTE), and it's one of the two major projects on which Apple is betting its future. If Newton flops, and if the Mac doesn't make the transition to the PowerPC chip, Apple may not survive the 1990s. At the very least, it could lose its seminal position as a leading-edge computer company. It may be years before PDAs match the sales volumes of desktop PCs, and whether Apple and others can wait that long depends partly on how many "early adopters" are willing to go out and purchase the first PDAs and fund the ongoing evolution. Nearly everyone agrees that those early adopters won't be the same people who are the eventual targets of PDAs. "We see the early adopters being PC users, but [sales] quickly moving to people who never considered buying a computer," says GeoWorks' Dougherty. Dougherty envisions a not-too-distant future when PDAs are as ubiquitous as radios. "The average household has seven radios, and most people don't even realize they have seven radios until they start counting their clock radio, their Walkman, the stereo system, the car radio, and so on. We see the same thing happening with devices like the Zoomer as they get down to calculator prices. And that will happen sooner than you think," he predicts. Dougherty estimates the bill of materials for a Zoomer-type device to be about $150, with the LCD being the most expensive component (about $30). Those costs will drop as volumes grow. Within two years, he says, a Zoomer could sell for as little as $300. That would put it in direct competition with personal organizers and pocket pagers. Dougherty points out that the current market for organizers is approximately 5 million units per year and that pagers are selling at a rate of approximately 2 million units a month. Dataquest projects that PDA unit sales in the U.S. will grow from 70,000 this year to 3.62 million in 1997. Like the primitive personal computers of the 1970s, the MessagePad and the Zoomer have an air of future potential, of something young that's just on the verge of making it big. It's only a matter of time before their handwriting recognition improves and new applications are invented for them. They fit well into today's busy lifestyles, and they're perfectly positioned to ride the rising waves of wireless communications and data superhighways. In 10 years, these unassuming little devices could conceivably become the dominant personal computer platform. Desktop PCs will still be around, as mainframes and minicomputers are today, but most "personal computing" will probably happen on PDAs. "There's a major intimidation factor with a computer," says Michael Reimer, chief operating officer of ParaGraph. "It has a scary-looking keyboard and a big monitor that stares back at you. The Newton is a very different kind of device: It's small, it's soft. It doesn't intimidate. It's like a Walkman." However, there are some naysayers, too. Michael Homer, Go's vice president of marketing, predicts that consumers will find the MessagePad and the Zoomer too expensive and disappointing. "I don't think the Newton and the Zoomer are viable machines. There's no magic going on here," he says. The future of PDAs isn't riding solely on the Newton MessagePad and the Zoomer. Both are promised to be only the first in wide-ranging families of products, and other contenders are poised to enter the race as well. Among the leading challengers will be the first PDAs to be based on Microsoft's WinPad and General Magic's Magic Cap. The millions of R&D dollars spent by Apple, Tandy, Casio, GeoWorks, Palm, Eo, Go, and other pioneers will not be wasted. Eventually someone will leverage the investment into a successful product. Let's just hope we don't have to wait until the twenty-fourth century for that to happen. Photograph: Apple's MessagePad is promised to be the first in a family of products based on the Newton architecture. Newton licensees are planning to introduce everything from smart telephones and fax machines to advanced cable TV converters. Photograph: Tandy and Casio collaborated on the Zoomer, a PDA that strongly resembles the Apple MessagePad but is built on a completely different hardware and software architecture. Photograph: The Eo 440 Personal Communicator is too big and too expensive to be a true PDA. However, it handles wireless communications in a complete and compelling manner. Table: WHAT MAKES A PDA? PDAs are defined as highly portable, easy-to-use computing and communications devices aimed at the mass market. To be highly portable, the unit must be easily held in one hand and able to be carried in a pocket. Long battery life is also an important portability factor; you don't want to worry about being near a power outlet all the time. Ease of use assumes no keyboard; most people don't type well. For now, a pen interface is the preferred means of input, although someday voice input might supplement it. To appeal to the mass market, PDAs must be low in cost, preferably $300 or less, and they must come with a good suite of general-purpose applications. Finally, PDAs must deliver seamless two-way wireless communications for fax and data access. (This table is not available electronically. Please see October, 1993, issue.) Sidebars:
Illustration: Newton Architecture: Newton Intelligence is object-oriented from the ground up. Most programmers will access MessagePad resources and write applications using NewtonScript. Illustration: The Layers of GEOS: The GEOS operating system is based on the familiar DOS architecture, but it does make use of some high-level objects for the sake of reducing code requirements for programmers. Illustration: Newton Data Storage: A store is the physical location of data storage, either main memory or a PCMCIA RAM card. A soup is a collection of related data. Entries are tagged frames of information in a soup. Soups automatically maintain indexes to their entries and can update or remove them on demand. Photograph: Newton MessagePad Motherboard Tom R. Halfhill is a BYTE senior news editor based in San Mateo, California. You can reach him on BIX as "thalfhill" or on the Internet at thalfhill@bix.com. BYTE West Coast bureau chief Andy Reinhardt, news editor Ed Perratore, senior technical editor at large Tom Thompson, and senior editor Michael Nadeau also contributed to this article. Letters / January 1994
I read with interest and fascination the excellent article
on PDAs ("PDAs Arrive But Aren't Quite Here Yet," October
1993) written by Tom Halfhill. At the same time, I confess
I was somewhat irritated by the absence of any reference
to Psion. Psion has manufactured and sold well over 1
million hand-held/palmtop computers, which is superior to
many of the other companies you cited. |