F E A T U R E S    Issue 2.09 - September 1996

Five Go Nuts in Cambridge

By Tom Loosemore

Acorn's crack techies thrive on new challenges, but their company doesn't. The Network Computer proves the point.



Peter Higgins is not pleased. Microsoft's senior vice president for desktop applications has flown half way around the world to preach Chairman Bill's Internet Gospel (2nd Edition) to a mixed bag of British journalists, only to discover that a new Messiah named Larry has nipped in first and left them all speaking in tongues about his new Network Computer. Higgins does not bother to hide his scorn. "I just don't get what it is you're so excited about. I mean, look who's going to make them. "Acorn". Acorn!!"

Acorn. To many, the very name is synonymous with failure, albeit failure of the heroic kind. Time after time during its rollercoaster 18-year life, the Acorn Computer Group has developed world-beating technology, only to fritter away its hard-won technical advantage through a mixture of bad luck, dreadful timing and almost unimaginable marketing ineptitude. The company has been described as the Sisyphus of the computer industry, forever rolling a stone uphill until it is within sight of greatness only for it to slide right back again. So when Larry Ellison, CEO of database giant Oracle, decided to entrust Acorn with the crucial task of producing the design for the new Network Computer ([trademark] Oracle) the choice appeared somewhere between the quixotic and the suicidal.

Except to some. Malcolm Bird, now CEO of Acorn's network computing division, thinks the decision always made perfect sense. "I think Oracle saw what Acorn had to offer them and thought Christmas had come early. We had exactly the technology they were after. And I think they also recognised that we've got some of the best engineers in the world working for us." Not an idle boast. Acorn has never really got the knack of making the right boxes, or selling them. But it does make boxes that work. And although Bird and everyone else at Acorn would come to see the Network Computer as something more, perhaps as a last, best chance at the big time, that was all Ellison wanted: a box that worked.

At least, that was all he wanted from Acorn. At another level, he wanted to change the world. Ellison, Oracle's visionary-in-chief, had long had a dream of a simple, cheap device capable of managing everyday computing tasks - email, word processing, spreadsheets, Web browsing - without the cost or the complexity of the PC. The Network Computer is the first physical manifestation of his dream. It's not aimed at those wanting to do desktop publishing or CAD; they need more power, which - in the short term, at least - only PCs can provide. Rather, Ellison's dream is of a true plug-and-play, mass-market connectivity device. Plug an NC into the power socket for electrons, into the TV for photons, into the telephone socket for bits, and away you go. No manuals, no setting up, no conflicting winsock.dll insanity. You plug it in, and away you play. Ellison wanted a computer that his mother could operate - ultra easy to use, and even easier to maintain and upgrade via the network.

Eliminating the local hard disk is one solution to the maintenance problem, both for Larry's mum and for corporate support departments, who spend an estimated £3,000 a year maintaining and upgrading the applications and data on desktop PCs. NCs have their operating systems and core applications etched into cheap-and-cheerful ROM memory chips; the device is useless without a network connection. All documents are stored on remote network drives, managed either (for home users) by an Internet service provider (ISP), or, for corporate users, by the IT department. And, critically, any NC anywhere in the world can be transformed into your personal machine merely by introducing a smartcard. Instead of hauling your portable around the world with you, you just take your card; it then manages your network connections, routing you to your own storage space on your own service provider's network server via local NC-aware ISPs. With the advent of the Network Computer standard, Sun Microsystems' oft-derided mantra "the network is the computer" would become day-to-day reality. But the closer you look, the more complex that vision becomes - and the bigger a gamble it seems.

The Great NC Gamble is not Oracle's. The world's number two software company has plenty of other full-uddered cash cows in the form of relational databases and server software. No, the gamble is Larry Ellison's. The NC concept is his submission for a place in history as The Man Who Broke The Wintel Oligarchy. History is not on his side; software companies have traditionally found the murky waters of hardware harsh and unforgiving, even Microsoft daring only to make a small if extremely profitable mouse. But Ellison seems to be jumping in at the very deepest end by attempting to engineer a whole new genus of computer. So why has he been willing to place so much faith in an oddball British company like Acorn?

A number of reasons. Acorn does have world class techies. It has a background in technologies very similar to that of the NC. And it has a pleasing lack of status. While Ellison was dashing round the globe on his NC blitzkrieg-cum-evangelical mission, Acorn's low media profile meant that it could work in relative obscurity - away from the prying eyes and pressures of Silicon Valley.

Oracle also hoped that Acorn's small size (1995 sales of £54 million, compared to Oracle's £1.8 billion) would allow it to be sufficiently responsive to come up with a product at breakneck speed. After all, it had done so before, if to little effect.

Back in the early 1980s, Acorn gained - and squandered - a massive head start in the emerging microcomputer business. It won the contract to build the million-selling BBC Micro, but only five years later failed miserably to exploit its own revolutionary RISC-powered, fully 32-bit Archimedes computer, the first - and for several years the only - genuine multi-media home computer. Now, Archimedes computers and their descendants constitute a tiny fraction of the European PC market; almost all sales are made to British schools and colleges.

And nearly a decade later, Acorn was one of the first to fall victim to the false prophet that was - and is - video- on-demand (VOD). It demonstrated that it could develop affordable, workable set-top-box technology, but a sizeable Cambridge trial, and the massive consumer indifference that followed, blew Acorn and its resurgent technology out of the water.

By early 1995 Acorn Group's owners, which include Olivetti Telemedia and Cambridge techno-Svengali Hermann Hauser, faced a dilemma. Losses for the previous year totalled over £4 million, and once the VOD bubble had burst they needed a new project to occupy their talented, if perennially underachieving, ward. They would take whatever was on offer. What they didn't expect was an offer quite as exciting as the one they got.

Something clicky

The gospel according to Ellison did not reach Acorn on tablets of stone, or even over the Net. It arrived on a video cassette. In October 1995 BBC TV's The Money Programme screened an interview with Larry Ellison. The interview stands up well as a piece of pure entertainment; Ellison was a man with a mission, and his message struck a chord with many in the computer industry, Acorn employees included. His vision of a pared-down, diskless, sub-$500 Network Computer was bold and exciting - but at that time, wholly impractical.

Such a computer needed a completely new design, one free of the proprietary architectures that have so hindered the development of today's desktop computers. Ellison had an opportunity to define a brand new set of rules, and to create a new culture in which his new device could flourish. The reference design for the Network Computer - the blueprint that defines what a Network Computer should be and how it should work - would determine every aspect of the machine, including networking, audio and video specs and smartcard design. It would be available to third-party hardware manufacturers, who would then build their own machines. But to make all this happen, Oracle needed to produce a working machine, one which could instruct and inspire other companies. And for that Oracle needed help from a serious hardware engineer.

Malcolm Bird was the man to offer it. Bird was one of the first to take the VHS of Ellison's Money Programme home from the offices of Acorn Online Media, a division set up in 1994 to exploit the projected VODboom. A well-built, imposing man in his mid-40s, Bird is an engineers' engineer. At our first meeting, shortly after the first Acorn NC prototype had been completed in an absurdly small time-frame, it required several minutes of prompting to get him to admit to being proud of what he and his team had achieved. "Mmm, I suppose you could say we're pretty pleased with ourselves." This is not false modesty - rather it is a case of misplaced English embarrassment when confronted with success. One of the junior members of his team had this to say of his boss: "Malcolm's a great engineer. He defines the problem, and then finds the least bad solution. He's brilliant at it. But, well, put it this way - he doesn't waste much time chatting with us about the footie."

Bird gives the strong impression that straightforward logic governs large chunks of his life, both at work and at home. In contrast to the cynicism and flippancy prevalent among those working for him, Bird treats frustration as just another problem to solve - step by step, working from the beginning. He is also an entirely arrogance-free zone. His grasp of technical minutiae is legendary, the respect afforded him total. At the launch of the Acorn NC, Hermann Hauser described him as "the man who knows more about how an NC works than anyone else on the planet."

Bird does not carry his many responsibilities lightly. During our meetings I set myself the challenge of making him laugh (ten points) or smile (five points). My final tally came to a meagre twenty-five. But despite his sober, rather earnest demeanour, all those on the second floor knew exactly where the buck stopped, and the freedom this afforded Bird's small team to take risks and speak their minds was to prove invaluable during the trials ahead.

The Online Media team is composed of a lot of old Acorn hands quartered in a two storey building on the outskirts of town - and a million miles from the city's courts-and-cloisters ideal. They have seen a lot of technologies come and go, and they are as cynical about new developments as it's possible to be. And yet, when their turn came to watch Ellison's BBC sermon, something clicked. His proposed Network Computer needed things Acorn already possessed in spades: a cheap, powerful processor in the form of the ARM 7500; a compact ROM-based operating system, RISCOS; TV compatibility. But best of all, Acorn had already built an NC. The ill-starred Acorn Online Media Set Top Box, designed as an affordable consumer appliance for the now-moribund VOD market, was an NC in all but name. Watching the VHS, Bird knew that Ellison had Acorn in mind when formulating his vision; it was just that Ellison didn't know it. Yet.

Larry does Geneva

October 1995. The Larry Ellison NC Roadshow hits Geneva for the Telecom '95 conference. After treating the assembled suits to his now-famous Microsoft-baiting NC stand-up routine, Ellison is buttonholed by Elsorino Piol, vice chairman of Olivetti. The gruff Italian has a notoriously impenetrable accent, but Ellison deciphers enough to realise that the two companies should talk further. Shortly afterwards, a video conference is arranged between Ellison, Andrew Laursen (the newly-appointed Oracle Network Computer VP), Hermann Hauser and Malcolm Bird, then MD of Acorn Online Media but soon to become CEO of the newly set-up Acorn Network Computing.

A week after the hour-long "knowing-me, knowing-you" video conference, Bird found himself on a plane to San Francisco carrying with him the latest Acorn Set Top Box. "I think they were fairly impressed," Bird recollects with characteristic understatement. They must have been. By the autumn of 1995, it was an open secret that Oracle had talked at length with several heavyweight computer manufacturers - Sun and Apple included - about the contract for the all-important first NC, the blueprint machine for a new generation of personal computers. Rumours were also circulating that Oracle also had sizeable internal team working on its own reference design.

When Bird returned to the UK, Ellison followed. By the time Ellison left, Acorn had a verbal agreement. Contracts and budgets could wait until later; both parties were confident that a bond of trust had been forged. For Acorn, it was a major coup. It would have its own machine built and on sale before anyone else. And even if it chose not to market its own machines, the royalties from the reference design alone could make Acorn a cash-rich company for years to come. The exact deal Acorn struck with Oracle is still under wraps, but if Ellison is right in predicting that the NC will outsell the PC by the end of the century, then Acorn's coffers will not be empty any time soon.

Sophie's world

As the company's chief scientist, Sophie Wilson is afforded near-guru status within Acorn. A Cambridge computer science graduate in the days when female ComSci graduates were decidedly thin on the ground, she accepted an invitation from Hermann Hauser to join what was then a fledgling computer sonsultancy way back in 1978. Socially awkward but extremely forthright, Wilson answers questions with a conciseness bordering on the curt. It's nearly impossible to hold a conversation with her when she is sat facing her radically customised Archimedes; she gives the impression - a misleading one, as it turns out - that she finds her beloved Archimedes both more interesting and more reliable than her fellow human beings. Underneath a layer of shyness she is incisive, dry and often caustically witty.

Wilson designed and built the motherboard for the Acorn System One, Acorn's first home computer. During the early '80s, she created the all-important instruction set for Acorn's first RISC chip, the antecedent of the hugely successful ARM family of low-cost, high-performance microprocessors. She is also the resident expert on the company's world-beating font anti-aliasing and anti-twitter technologies, which both employ the same clever optical trick to make using a television as a computer screen altogether less painful than it might otherwise be. Unsurprisingly, younger and less experienced staff regard her with awe. Bird describes Wilson as a walking encyclopaedia, the embodiment of a vast body of knowledge culled from Acorn's previous adventures. This is not just technical know-how; Wilson has seen many Acorn projects run into difficulties due to confused, indecisive management and bloated teams, and she was not prepared to see the NC project fail for the same reasons.

Even before an official deal had been struck, Wilson sent a prophetic email to other members of the team, detailing the shortcomings of the NC's technology and voicing reservations about the machine's progenitor. "Larry Ellison is a very hands-on person. He will become involved in many ways, and will leak to the press." Yet she had other, more positive messages to deliver. Among the many implications of the NC deal she elucidated, two words stood out: "World Stardom".

Following their reconnaissance mission, Ellison left Andy Laursen behind to talk tech with Sophie Wilson. By the time Laursen returned to San Francisco in mid-November, the pair of them had thrashed out a draft NC specification. Wilson had "guesstimated" a timescale of just nine weeks to produce a working prototype, which she thought eminently manageable; "Hey - the BBC machine prototype was built in six days!"

Nonetheless, they needed to move desperately quickly if they were to have a working NC in time for Ellison's first mediafest in mid-February 1996. Building a new computer from scratch is a non-trivial undertaking, even with many of the components at hand. The hardest parts of the process - planning the layout of the mother board, designing the case, writing the software - had to be done from scratch, and the July manufacturing date looked worryingly close.

Malcolm Bird spent the final two weeks of November 1995 assembling a small team of old hands. Colin Watters, a quiet, hamster-like man in his early 40s, was appointed as the project manager. Acorn has been part, but not all, of his life for eleven years; an electrical engineer by training, he previously managed the successful Archimedes A5000 project to produce a low-cost educational version of the Archimedes. Also an avid radio-controlled model aircraft pilot, he recently flew for Great Britain, an honour he was extremely coy about revealing. The core team also consisted of Steve Cormie, a 29-year- old software expert with a sharp eye for detail and an even sharper wit, and Richard Sharples, an enthusiastic, jovial eccentric in his early 30s whose remit covered motherboard design and all things CAD. They were ready to go.

"It's a design problem"

After two manic weeks spent sourcing parts and finalising the exact prototype specification, the team hit a large - but not entirely unexpected - snag. Ellison had set his heart on a radical, imaginative design for the NC's casing. He had commissioned Seattle-based Frog Design, designers of his favourite Bang and Olufsen HiFi equipment and of the original Macintosh computer, to create a case that would capture the imagination of a public bored with standard-issue grey plastic PC boxes. But when the Cambridge team saw Frog's initial drawings, its members issued a collective groan. The cruciform shape was certainly eye-catching, but it would be a nightmare to manufacture.

Oddly, designing the case is one of the most difficult parts of creating a new computer. Manufacturing a shell which both fits together and can be manufactured cheaply is in itself hard; the fact that its various elements look like they should all fit together in 2D view does not mean that they will do so when the mock-ups are actually built. To make things worse for the Acorn team, Frog Design's planned casing would not leave enough room for the printed circuit board (PCB).

Worried by this, Bird made a bold move. He commissioned a local two-man design company, Cambridge-based Design Edge, to come up with an alternative within 48 hours. And it did. Design Edge's planned casing would win no prizes for innovation or daring, but Bird was nonetheless pleased; it was tidy, it could be built cheaply and in a hurry. To appease Ellison, Richard Sharples persevered with Frog Design's proposed cas- ing, finally convincing Oracle that, pretty though it undoubtedly was, there was simply no way to fit the PCB inside the case without resorting to expensive right-angled daughterboards and consuming more precious time.

Five get sloshed

For Malcolm Bird, at least, the first week of January 1996 signalled an end to some of the uncertainty surrounding the project. The details of the contract between Acorn and Oracle had been finalised, and Sharples had completed the PCB design and had sent it off to be manufactured. The final drawings for the case were in the hands of Peter Manning, a professional model maker based in St Ives who had the fiendishly difficult job of transforming the subtle curves described on the Design Edge drawings into an exact 3D mould ready for casting. Within two weeks Manning delivered the goods, on time, and Cormie shipped the first batch of system software and demos over to the US. Unbelievably, they were actually ahead of schedule; although Watters never expressed as much to Bird, he was confident that they could have a working prototype ready two weeks early.

In the third week of January, a five-man delegation arrived from Oracle. Bird kept them well away from the team while the newly arrived circuit boards were fitted and tested. "We knew it would work okay," Watters says. "Our main concern was that Richard had made some idiotic oversight and the bloody thing wouldn't fit together."

In the event, the case didn't merely fit; it fitted like a dream. The Acorn team had designed and built a computer in seven weeks, and the astonished Oracle delegation returned to San Francisco carrying the world's first NC in a Tesco bag.

The team celebrated with a Friday afternoon visit to the pub, where, Colin Watters freely admits, "we all got a little bit sloshed." But the cele- bration was short-lived. As a reward for their efforts, Oracle brought the whole schedule forward by two weeks as Ellison leaked a story to the press that the NC would ship in March. As Malcolm Bird puts it, "The March shipping date was a bit of a joke, really; it was both a slap on the back and a slap in the face."

Then, Oracle suddenly wanted not one but twenty working NCs for the mid-February mediafest. The team immediately ran into component-sourcing problems; whereas previously the team had had no problem blagging five or six component samples from a supplier, things got a lot more complicated when it began asking for fifty. This time, things went right to the wire; at 2pm on February 16th, there were 20 NCs in the building. Unfortunately they were still in pieces, and a courier was due to arrive at 5pm to whisk the completed machines off to San Francisco. Turning the pieces into machines was simple enough - the whole office was put on assembly-line duty - but the real headache involved filling in the mass of customs forms required to import RAM into the US. Somehow they made it, with only minutes to spare; Sod's Law being what it is, the customs forms were never checked.

By the time of my next visit in mid-April 1996, some far more serious problems had emerged. The atmosphere in the office had undergone a seismic shift; the excite- ment had passed, and the team was struggling for motivation. March had come and gone and there was no NC even close to being manufactured. Worst of all, the jewel in the NC's crown - the ARM7500 processor chip - had been found wanting.

Battle stations

The ARM7500 is the pocket battleship of processors, a little-known miracle in silicon. Roughly equivalent in power to an Intel 486-66, but around half the size and price, the chip requires astoundingly little power; Colin Watters claims that the 7500 could theoretically run off the waste heat generated by a Pentium chip. Indeed, the ARM7500 is actually too power-efficient, with an extra power-wastage resistor required to ensure that the AC-DC transformer transmits enough current to ensure a steady voltage. Miracle though it was in its own right, both Acorn and Oracle had come to the conclusion that the 7500 lacked sufficient horsepower. Bizarrely, Sophie Wilson had known this since the earliest days of the project - her original email to the team admitted that the 7500 is "old, has slow memory, slow CPU and no [floating point] hardware; it can't drive cheap colour LCDs." And she had anticipated pressure from Oracle to improve the chip.

The LCD problem was especially serious, as LCD compatibility was essential for portable versions of the NC family, yet no work had been done on designing the 7500's successor. Belatedly, the decision was made to bolt a floating point processor to the ARM7500, speed up its on-chip cache, increase its overall speed to 40MHz and call it the 7500FE. There was, understandably, an air of impending panic. Even the usually phlegmatic Watters was concerned. "Yeah, I suppose it is, um ... rather high risk.... But if it's got to be done, then we'll knuckle down."

As if that wasn't enough, the browser software, now ready to be incorporated into ROM, looked decidedly ropy. A derivative of the unpopular ANT browser, it still lacked support for frames and tables, features which were now integral to many Web sites. Steve Cormie remained grimly adamant that the final version would sup- port these as well as Java, but such an upgrade was a job for a whole team of software engineers, not one tired-looking individual.

Why not just license Netscape? I asked Cormie, receiving in return the kind of withering look that software engineers reserve for those asking particularly dumb questions. "How much memory does the latest version of Netscape need to run? 4Mb? 8Mb? Whatever, it's more memory than will be in the whole bloody machine, ROM included. Who wants Netscape plug-ins anyway? Nah, ANT will have to do."

Such an attitude betrayed one of the main weaknesses of the team as a whole. The Web, in all its ever-expanding, ever-mutating glory, was perceived as a childish distraction, an entertaining plaything for those unable to hack it at the cutting edge.

Surfing was simply not their bag; a specialised Web-browser is a machine that they would never dream of using themselves. Sites designed with Netscape in mind looked abysmal when viewed with through the primitive, out-of-focus eyes of the ANT browser; technically wonderful though the Acorn NC is, it is certainly not the ultimate Web machine.

At the end of May, the Oracle NC partnership was announced in San Francisco. Set against a room- ful of other NCs cobbled together around the early specification by Oracle's other partners, Acorn's machine - now branded as the NetStation - showed its peers a clean pair of heels, even without the 7500FE chip. And Steve Cormie and the ARM techies had managed to squeeze frames and tables into their tiny ROM-based browser. The team was in Cambridge, trying to get an NC into the shops by Christmas.

Let's do launch

On the day of the UK launch, June 18th, the 7500FE machine still isn't working. The launch goes ahead regardless, and it is immediately clear that the attitude of those charged with marketing the NetStation contrasts sharply with that of the team that built it. The event is a sham, an exercise in smoke and mirrors, as Hermmann Hauser and assorted cronies attempt to convince a deeply sceptical audience that this small, dull black box will challenge the TV as the chief entertainment device in the nation's living rooms.

Web hype is now so frenzied that the NetStation will quite likely fly off the shelves when it hits the high streets in September. But what people will be buying is a half-finished gimmick, not the device that Larry Ellison described on The Money Programme. Later versions of the NC - whether they're made by Acorn or by Apple - may fulfil Ellison's true vision, but for the launch, the NetStation's marketeers have submerged the broader concept in a blind act of faith in the Web as a killer consumer app. Sadly, sales are likely to fall away before the hype does. Until there is serious bandwidth into the home, the NC is a non-starter as an entertainments device; that will come in time, but not by this Christmas.

The great tragedy of all this is that having set the standards and built the first serious machine, Acorn will never be a serious competitor in the market for network computers. As it tries gamely to flog a turkey to people who don't yet need one, its competitors will clean up in more lucrative corporate markets where bandwidth is already available, but to which Acorn's marketing machine has little or no access.

Does this worry the members of the Cambridge team? Not really. They've made a good machine, and they'll make others. It may rankle that the the inept antics of the marketeers have denied them the "world stardom" Sophie Wilson so eagerly anticipated, but so long as other meaty challenges remain to be tackled, they'll be happy at heart. Aside from the 7500FE chip's inevitable teething troubles, Wilson has already coaxed one of the new-generation 150MHz StrongARM chips into driving an NC, and other work is in progress. It's as if the marketing fiasco simply confirms that their own status quo has not been disrupted. All in a normal year's work at Acorn - and a box that works to prove it.

Tom Loosemore is a section editor at Wired.