Dead End: Palm and the PDA

Abstract

Palm Computing built the first handheld computer that ordinary professionals actually bought: the Pilot 1000, March 1996, $299, more than a million units in eighteen months. It beat the Apple Newton by doing less, and by making the user learn a stroke alphabet instead of making the machine learn cursive. Over the next fourteen years Palm was bought by US Robotics, absorbed into 3Com, floated in an IPO that briefly made its parent company worth less than nothing, split into two companies, reunited with the breakaway startup its own founders had built, and sold to HP for $1.2 billion. Thirteen months after that deal closed, HP shut the hardware down. webOS, the operating system Palm built to answer the iPhone, now runs LG television menus.

Jeff Hawkins and the Plywood Prototype

Jeff Hawkins founded Palm Computing in 1992 after GRiD Systems, where he had designed the GRiDPad, an early pen tablet sold mostly to industry and the military. His interest was the use case rather than the technology: replace the paper organizer a professional carried, not shrink a computer.

Palm’s first product was not a device. It was the software inside somebody else’s: the Zoomer, a 1993 PDA built by Casio and sold by Tandy, for which Palm supplied the personal-information applications and the PalmPrint handwriting recognizer. The Zoomer sold poorly. Palm’s second product was Graffiti, announced on 19 September 1994 and sold as an add-on for other companies’ machines, including the Casio Zoomer, the AST GRiDPad 2390, the Sharp ExpertPad and the Apple Newton. Palm’s stroke alphabet was on sale as a fix for the Newton’s handwriting problem two years before Palm shipped hardware of its own.

Graffiti inverted the recognition problem. Instead of teaching a computer to read human handwriting, it defined single-stroke characters close enough to ordinary letters to be memorable and distinct enough that no two could be confused. A user could learn the alphabet in about twenty minutes, after which entry was reliable. Hawkins’s argument was that people already learn to touch-type, so learning a writing system was a reasonable price for a machine that never guessed wrong.

The other requirements were defined by exclusion: nothing larger than a shirt pocket, nothing slower than one second from button to information, nothing priced above what a professional would buy without asking a purchasing department. The size test was run with a block of wood cut in Hawkins’s garage to fit his shirt pocket, carried for weeks and tapped with a whittled chopstick whenever he would have used the real thing. The exercise settled both the dimensions and the feature list: calendar, addresses, to-do list, memo pad, and nothing else. The wooden block is in the Computer History Museum’s collection.

The Pilot

US Robotics bought Palm Computing in 1995 for $44 million, which supplied the manufacturing and distribution Palm lacked. The Pilot 1000 ($299, 128 KB) and Pilot 5000 ($369, 512 KB) went on sale in March 1996.

The Pilot ran a month on two AAA cells and turned on instantly, because it never really turned off. Four hardware buttons launched Calendar, Contacts, Tasks, and Memo directly, with no menu in between. HotSync, a cradle wired to the desktop’s serial port, reconciled both sides in under a minute in either direction: change an appointment on the desktop and it appeared on the Pilot, and the reverse. For anyone carrying a Filofax the trade was obvious, and US Robotics sold more than a million units in the first eighteen months, outselling every other handheld on the market.

The name lasted less time than the product. Pilot Corporation of America, which makes pens, sued over the trademark. The second generation, launched 10 March 1997, shipped as the PalmPilot Personal ($299) and Professional ($399); after the case settled in 1998 the word Pilot disappeared entirely, and the machines became the Palm III, Palm V, Palm VII.

3Com acquired US Robotics in June 1997 and inherited Palm with it. 3Com treated the handheld business as a peripheral line rather than a platform, and Hawkins, Donna Dubinsky (Palm’s chief executive), and Ed Colligan (marketing) found their expansion plans underfunded.

Graffiti in court

Xerox sued US Robotics in April 1997 over Graffiti, claiming it infringed a PARC patent by David Goldberg covering Unistrokes, a single-stroke input alphabet developed at PARC in the early 1990s. The case ran for nine years through infringement findings, an appeal, and a 2004 district-court ruling that the patent was invalid. Palm shipped Graffiti 2 in 2003 to route around the dispute, and in 2006 paid Xerox $22.5 million for a paid-up licence to Unistrokes and two other patents plus a seven-year truce. The stroke alphabet that made Palm’s reputation was, legally speaking, contested for most of the company’s life.

The Software Market

Palm OS was small, single-tasking, and documented. The SDK was free, applications were written in C, and there was no approval gate. By 2000 the catalogue ran to more than 10,000 titles: drug references carried by physicians on rounds, GPS navigation using external receivers, translators, Bible software, legal references, games. Palm ran an application market eight years before the App Store, with the difference that nobody took 30% and nobody could be removed from it.

Hardware followed. The Palm V (February 1999, $449) was the design statement: anodized aluminium, lithium-ion rechargeable, thin enough to disappear in a jacket pocket. The Palm VII (May 1999, $599) put a flip-up antenna on a PDA and sold wireless access to clipped-down web content over the Mobitex network for $14.95 a month, years before phones did anything comparable.

Then came the IPO. Palm was floated out of 3Com on 2 March 2000, priced at $38, opened at $150, and closed its first day at $95.06. At that price the 95% of Palm that 3Com still held was worth more than all of 3Com, which meant the market was pricing 3Com’s entire remaining business, its networking products, its cash, at roughly minus $22 billion. Economists Owen Lamont and Richard Thaler used the episode as the cleanest available evidence that the dot-com market could not add and subtract. Palm’s shares lost about 90% of their value within a year.

Handspring

Hawkins, Dubinsky, and Colligan left in June 1998 to found Handspring. The disagreement was about direction: they wanted to build a device that was also a phone, and 3Com did not want to be in the phone business.

Handspring licensed Palm OS from Palm and shipped the Visor on 14 September 1999, starting at $149, with a Springboard expansion slot that accepted modules: memory, modems, MP3 players, cameras, and eventually a phone radio. It undercut Palm’s own hardware and sold well.

The Treo 180, shown in October 2001 and shipped in January 2002, was the machine Hawkins had left to build: Palm OS, a thumb keyboard, and a GSM radio in one pocket-sized body. The Treo 600, announced in June 2003, made it good, with an integrated keyboard, a camera, and cellular data.

By then the two companies were merging back together. Palm bought Handspring in an all-stock deal worth about $169 million, announced 4 June 2003 and completed that October. The same transaction split the operating system away: PalmSource was spun off as a separate public company, and the hardware business, merged with Handspring, became palmOne. Palm’s phone maker and Palm’s platform were now different companies with different shareholders, at exactly the moment the smartphone market started to matter. ACCESS Co. of Japan bought PalmSource in late 2005 for about $324 million; palmOne bought back full rights to the Palm name in May 2005 and reverted to the name Palm that July.

The Platform Trap

Palm OS had been built for an organizer: single-tasking, no memory protection, no expectation of a radio. Smartphone work (background sync, a real browser, cellular state, push email) needed a rewrite that nobody was in a position to fund, since the platform now belonged to ACCESS.

Palm hedged in three directions at once. Palm OS Garnet kept the Treo line shipping. ACCESS built a Linux-based successor, the Access Linux Platform, which never appeared in a significant product. And in September 2005 Palm announced the Treo 700w, running Microsoft’s Windows Mobile, which reached Verizon in January 2006. Palm was now selling hardware for a competitor’s platform, with margins to match.

The Treo 650, 680, and 700 series kept Palm in the market without moving it forward. BlackBerry held enterprise email, Windows Mobile had the broader application catalogue, and Nokia’s Symbian led global volume.

Myth: Palm’s CEO said the “PC guys are not going to just figure this out”

Ed Colligan’s dismissal of Apple’s rumoured phone, made at a Churchill Club breakfast in November 2006 in answer to a question from New York Times reporter John Markoff, is quoted almost everywhere in the form “We’ve learned and struggled for a few years here figuring out how to make a decent phone. PC guys are not going to just figure this out. They’re not going to just walk in.” When the Computer History Museum’s Churchill Club recordings were consulted in 2022, what Colligan actually said was: “The PC guys are not going to just, you know, knock this out. I guarantee it.” The famous wording is a contemporary paraphrase that hardened into a quotation. See Myths and Misconceptions.

webOS

Apple’s iPhone announcement on 9 January 2007 reset the requirements: a real browser, capacitive multi-touch, and hardware and software designed together. Palm’s answer, webOS, was shown at CES on 8 January 2009. It ran a Linux kernel under an application layer written in HTML, CSS, and JavaScript, so an application was a web application in a native frame. The Cards interface represented each running program as a card in a horizontal stack that could be flicked off-screen to close it, which iOS and Android later borrowed in substance. Synergy merged contacts and calendars from Gmail, Exchange, Facebook, and the phone itself into single entries. Just Type searched everything from the keyboard before any application was open.

The Palm Pre launched on 6 June 2009, exclusive to Sprint, with a slide-out keyboard and a plastic body that felt cheap next to an iPhone. It also synced with iTunes, by identifying itself over USB with Apple’s vendor ID. Apple broke that in iTunes 8.2.1; Palm restored it, complained to the USB Implementers Forum, and in September 2009 the forum ruled for Apple and told Palm that borrowing another member’s vendor ID was itself a violation.

The catalogue was the killing problem. webOS had about 1,000 applications in January 2010 and 4,000 by that September. Apple had announced 100,000 in November 2009 and passed 300,000 during 2010. A buyer choosing a phone in 2010 was choosing an application library, and Palm’s was two orders of magnitude short.

HP

Hewlett-Packard announced the purchase of Palm on 28 April 2010 at $5.70 a share, about $1.2 billion, under chief executive Mark Hurd, and closed it on 1 July. Hurd resigned that August. His successor Léo Apotheker kept the stated plan of putting webOS across HP’s product lines, phones, tablets, and printers, then reversed the company’s direction.

The HP TouchPad tablet shipped on 1 July 2011 at $499, the same price as an iPad, with better software than its reviews suggested and nothing like the iPad’s application library. Forty-eight days later, on 18 August 2011, HP announced it was ending all webOS hardware, as part of the same strategy shift in which Apotheker proposed spinning off HP’s PC business. Remaining TouchPads were dumped at $99 and sold out within days, which established that the device had been priced roughly $400 above its market.

HP sold the webOS team and code to LG Electronics in February 2013. It has run LG’s smart televisions ever since, which is the largest install base webOS ever had.

The webOS diaspora

Matías Duarte, who led webOS design at Palm, joined Google in 2010 and became the designer behind Android 4.0 Ice Cream Sandwich, the release that gave Android a coherent visual language. Card-based task switching turned up in iOS 7. The interface ideas outlived the company that shipped them.

Dead End: Inventing the Category, Missing the Wave

Palm was right three times. The Pilot was right that a pocket organizer beat a pocket computer. The Treo was right that the organizer and the phone were one device. webOS was right that a mobile operating system could be built out of web technology, which is roughly what every cross-platform framework since has assumed.

What Palm never had was control of its own stack at the moment it needed one. The 2003 restructuring put the hardware and the operating system into separate companies just as the smartphone transition began, and Palm spent 2005 to 2008 building phones on a platform it did not own, then on a competitor’s platform, then on a new one written from scratch under time pressure. Apple shipped hardware, operating system, store, and carrier deal as one negotiated package. Palm shipped a good phone on one carrier with a thousand applications.

The rest was arithmetic. Sprint exclusivity capped the Pre’s audience while AT&T’s iPhone exclusivity ran to 2011 across a far larger base. HP paid $1.2 billion for the platform and abandoned it inside of thirteen months rather than fund the years of losses that building an ecosystem would have cost. Inventing a category and surviving it turned out to be separate problems.