Symbian's Secret History: The battle for the company's soul
How Nokia took charge, and never let go
So what really happened?
Various reasons have been put forward over the years: the market wasn’t ready, the silicon wasn’t mature, or the networks weren’t in place. Executives cite several others here. Some are interlocking reasons and few are mutually exclusive.
The reality that hit both Symbian engineers and their customers building the phones was that putting Epoc into a complex communication product was much harder than people realised. In particular, moving the OS to new, custom hardware – a job called "base porting" - was problematic. There was no market of standard off-the-shelf integrated components, as there is for smartphone manufacturers today - although this would rapidly evolve. In a couple of cases, such as Motorola's MCore and the processor for Ericsson's first R380, the chip had no MMU (memory management unit).
So base porting was truly virgin territory - and because the phones were an order of magnitude more sophisticated than anything on the market, their complexity added more issues. This may seem so obvious in retrospect, it's almost a tautology. Yet we forget how relatively crude the hardware was then - it was striking in researching this article how much we take for granted.
Base porting was always going to be a challenge, then. But it was additionally problematic, according to senior Symbian staff, because the Symbian partners couldn’t agree who should do it.
“Nokia and Ericsson told Symbian that we should leave that task to them. So Symbian OS always had holes at that level,” recalls one former executive. “Over time, these holes caused more and more problems to new licensees - and even to the old ones. In retrospect, this may have been the single worst decision about the design of Symbian OS.”
Another reason was that the silicon was genuinely immature. Symbian engineers spent many man years of work coding around bugs in the chips. (See: Waiter - these chips are undercooked).
Delays tended to cause more delays.
Against this background was a dramatic realignment of the industry. The telecomms bubble had burst, impacting companies across the industry. The phone manufacturers biggest customers, the operators, had badly miscalculated the appeal of WAP - then paid billions for 3G licences. Three of Symbian's big four shareholders - including Ericsson and Nokia - were large suppliers of telecomms equipment to the operators, as well as suppliers of handsets. Ericsson also experienced a dramatic fall, and spun out its handset business into a partnership with Sony in 2001. For all its consumer electronics know how, and clever designs, Sony had failed to make any headway with its GSM handsets. None of these moves made customers look at their risky, expensive Symbian smartphone projects with confidence.
Much of the delay in this period can be explained by manufacturers failing to bring projects to market - there was no shortage of activity. Sony had invested in a Symbian phone with a 3D display that impressed all who saw it - but it never reached the stores. Panasonic never brought its smartphone to market, and only ever offered a couple of lacklustre models much later in 2005. Kenwood's phone was also canned.
Once initial targets were missed, the major shareholders insisted upon introducing micro-management, which also slowed progress further. Measuring and monitoring, as in any bureaucracy or company lacking leadership, had displaced development as the focus.
Milestone or Millstones?
One decision that was hotly contested was Symbian's decision to deliver regular milestone releases every six months. This was supposed to assure customers that the new venture was a dependable supplier. But it had several consequences.
"Releases bore no relation to the phone manufacturer's schedules. So releases would be made that nobody wanted because either everyone was busy making a phone on the last release or they weren't ready to start a new phone yet. Some releases simply disappeared because they had no customers," notes a senior engineer.
Phone manufacturers wanted the new features in the older stable versions they were working with, so the features would have to be backported - an unpopular job at Symbian. Sometimes milestone releases had to drop almost everything that had been promised. And with a distant window, the engineers went crazy in the sandpit.
"It gave carte blanche to the software engineering teams to just go away and break everything after each release, because they didn't have to have a complete working system until 6 months in the future," he adds.
Today we see the prolem hasn't really gone away. Phone manufacturers chafe at Google's impressive, rapid release schedule for Android - and insist on backporting features to their older stable versions.
As the UI air wars took place, back on the ground Symbian gradually became a custom product design house, working closely with the owner/partners to bring the phones to market. In the early years revenues from consulting far exceeded revenues from royalties. (Royalty revenues actually fell in 2001).
"The feeling at one point was that Symbian couldn’t deliver anything, that nothing was going to come out of it,” recalls Gretton, who had built up Psion’s software team from scratch after Symbian was formed. “The computer side got so frustrated we considered using alternative OSes."
And this in turn, impacted Symbian’s bid to continue its UI work.
"In the middle of this period", recalls a senior engineer, "the tech bubble burst. And suddenly all the limitless money that Symbian management had been promising (and people in the company had been spending) dried up. Nokia had wanted its "Calypso" phone, the 7650, on the market by the middle of 2001. It cranked up the pressure, and having thrown so much into its own user interface, quietly decided it wanted to renegotiate one of the untouchables of the Symbian agreement - the royalty schedule.
Roger Nolan, who set up Symbian’s technical consulting operation, recalls:
"Nokia decided that 'you're having trouble delivering the base stuff - what are you doing with this other stuff?'"
He remembers one executive board member from a large Nordic handset manufacturer banging the table during a meeting, shouting:
"We gave you the fucking money, now give us the fucking software."
Waiter - these chips are undercooked
What does an engineer do when the erratum sheet to a chip is longer than the chip's technical reference manual? Most would be tempted to slip quietly out of fire exit, and not start screaming until they'd reached the car park. But this kind of problem faced Symbian engineers daily as they grappled with immature and poorly documented silicon.
"There was a fundamental difference in outlook," recalls one former Symbian engineer, who was faced with making the OS work on new and foreign hardware for its phone makers.
"When Symbian said 'reference', it meant product-quality, something that you could take and use in a phone and then add to. By 'reference' the semiconductor companies (semicos) had always meant something rough and ready that was just enough to demonstrate the principles of getting the hardware to do something, you were supposed to use it as a guide to creating a proper version.
"Semicos like Texas Instruments were happy with something that just booted and showed off the headline features of the chip, such as video playback. They didn't care about finishing it to product quality, or filling all the holes, or supporting it, or making it portable - to them that was stuff that their customers were supposed to do."
Few chip companies were even interested in a market that didn't exist, but one that was - and later reaped the reward - was Texas Instruments. TI supplied custom chips to Nokia, while Ericsson designed its own, and had a close relationship with VLSI, which had recently been acquired by Philips. For much of this period, Ericsson came out better.
"TI had decided to save money by designing their own data cache instead of licensing the ARM one," he recalls. "So none of the revisions of the OMAP 15xx that were delivered ever had a reliable data cache. On the same chips, the technical reference section on the built-in real-time clock advised you not to use it because it used too much power."
Ericsson would sell its semiconductor business off in 2002, but continued to design its own silicon with VLSI (now Philips). The NX processors it designed for its P800 phone, up to its P1i, were very efficient - and relatively clean.
Nightingale: a tactical retreat
"Nokia thought they’d created a monster – now let's take it away,” says Christensen.
Nokia succeeded in crushing the Symbian UI vision, but not everyone attributes nefarious motives to the Finnish giant.
“Nokia’s vision of the future was as sound as anybody could predict. So as soon as we lost our battle for the UI, all the others went back to their flavour of week kind of thing,” Randall remembers.
"I don't think anyone was being dishonest. We didn't realise how controlling Nokia would be, when push came to shove," says Roger Nolan, who managed the Symbian technical consulting operation. "A lot of the time the board that were running the company [the Operations Board] thought they had a lot more autonomy than they did."
Another Psion designer, Matt Millar, says the politics hampered the vision.
“Symbian knew what consumers wanted, what software they wanted. But it got pulled apart by competing views – it could never execute on Symbian view.”
Symbian board sounded a tactical retreat from user interfaces - codenamed Nightingale.
"We were having difficulty staffing all the licensee projects. Only Quartz and Crystal survived as living code as UIQ and Crystal respectively," Nolan remembers.
Symbian and Nokia agreed to allow a startup - Matt Millar's Mobile Innovations - to take control of Crystal and consult on the forthcoming Nokia Communicator.
"Of course, it was tactical for Symbian but strategic for Nokia - they ensured that they were the only people who could cheaply produce Symbian phones."
Nightingale was a landmark decision – it ensured that while Symbian would remain a powerful industry force, it wouldn’t set the agenda. Most now see it as inevitable – but not everyone. Christensen today believes a more entrepreneurial Symbian would have called Nokia’s bluff, and gone outside the club of original owners for investment.
“Where would Nokia have got another OS from?”
After Nightingale, Symbian could never quite decide whether it should keep a foot in the business of designing and licensing UIs. It sought to spin out Quartz as a joint venture, jointly owned by Sony Ericsson (as it had become) and Motorola, but CEO David Levin pulled it back in, allowing it to continue as an independent subsidiary. And so it did, until Sony Ericsson acquired it in 2006.