The Symbian OS (779886), страница 8
Текст из файла (страница 8)
Of course Colly was right, but the fact is that it’s just too hard to do theright thing. So I walked in one Monday morning and checked out a releasenote, this would have been October 1995, it just read ‘As agreed, changed allTUint to TInt.’ Of course we kept TUints for flags and such – but as fornumbers, that was how that debate got closed.
And Colly’s ‘agreement’, whenit came, was characteristically unilateral, announced through the release noteafter a gruelling weekend’s work which couldn’t really be automated – Collyreally had to check every change manually.The relentless development pace and constant project pressure werehard, too. Peter Jackson, who these days is responsible for Symbian’ssoftware configuration management systems, remembers the approach toproject management, as directed by Bill Batchelor, with mixed feelings.Peter Jackson:Bill Batchelor liked getting his hands dirty, but then he became project managerfor Protea and so he had a dual nature.
He was passionate about the rightthings, but at the same time he didn’t like it when you told him how long itwould take to do ‘the right thing’.But the company was vibrant and small and people didn’t actually mindhacking away for all hours of the day and night to achieve the end goal.Bill’s over-optimistic project plans were just part of that mix. He would cajoleTHE BEGINNINGS OF SYMBIAN OS25you into committing to something that was impossible, and you’d do yourbest to achieve it, and then eventually there would be this undercurrentwhere everybody knew it was totally absurd, but no one was going to say it.Eventually it would all come out. And then there would be another planningiteration and the same thing would happen again.The big practical problem with the Protea project, one which caused asuccession of headaches for Geert Bollen, was the lack of real hardware.Software development started well ahead of the availability of any prototype hardware but even by mid-1995, when the software project was infull swing, the device prototypes were still not ready.Geert Bollen:The on-the-metal version of the kernel was started and delivered after I arrivedand Colly Myers assembled a team for that.
Before that Colly had been aone-man band. The GNU tools at that time were coming on. I had someinvolvement in that but they were still a long way from being rolled out.Andrew Thoelke is another veteran who joined in March 1994 andis now the Chief Technology Architect at Symbian for the base servicesand kernel layers of the system. In the absence of hardware prototypes,built code was run on PCs using an emulator layer which mimicked a fullsystem by mapping low-level operating system calls to their Windowsequivalents, essentially the same approach used by Symbian OS developers today in the first stages of development, before moving to hardwaretargets.Andrew Thoelke:Down in the base team, not having hardware was a problem, so the systemwas first brought up on x86 as a hardware port before it was ever brought upon ARM.
In the original kernel architecture, probably 40% or 50% of codeis shared with the target, but there’s still vast amounts of kernel code whichis target only, all of the scheduling and threading, the interrupt model, thedevice-driver model, so all of that needed to be done with a real target. Sothey used a 486, they basically built an 8386 port of the system first, becausethat brought online another 40% of the kernel code. Obviously there was stillARM-specific hardware code and a different MMU and all that sort of thing.But it was actually much less work when hardware did become availablebecause they had already got a generic kernel mostly working.Whatever the problems, there was no doubt in anyone’s mind thatwhat they were creating was special. Martin Budden, now Chief System26THE HISTORY AND PREHISTORY OF SYMBIAN OSArchitect at Symbian, was a veteran of the two 16-bit projects beforemoving onto the EPOC project.
He puts it very simply.Martin Budden:I came to the company because I wanted to do something that was exciting.As soon as I saw what Psion was doing, I just knew that was what I wanted todo.Looking back, Psion’s timing was good; it had judged the momentperfectly.Martin Tasker:Psion, like many companies then and not just in Britain but elsewhere, hadachieved success by innovating according to rules which nobody had everwritten. It just did its own thing and it found a niche in the market.2.5 The Mobile OpportunityWhen the Psion board decided to spin off its software division, whichat that time numbered around 70 engineers, it was effectively a publiccommitment to a software-licensing strategy.It is clear that a number of different options were considered.
Therewere rumors at about that time that Psion had considered buying Palm. Apossible purchase of Amstrad got as far as due diligence. The backgroundis revealing. For the Psion board, the real target seems to have been aDanish phone-making company, Dancall, which Amstrad had previouslybought and absorbed into its empire.
Thus, buying Amstrad would haveenabled Psion to become a phone manufacturer. This indicates veryclearly the direction in which Psion was pressing at that time. In theevent, Psion did not buy Amstrad and Dancall was eventually soldto Bosch, before being sold on to Siemens. Much later, it was theformerly Dancall site at which the Siemens Symbian OS phone, theSX1, was developed. Still more recently, the site has been sold on toMotorola.Psion, of course, did make its move into the phone market, but in aquite different direction.
It was a visionary move and one for which thecompany founder David Potter deserves enormous credit. There wereother visionaries too. In particular, Juha Christensen, Psion Software’sBACKGROUND TO THE FIRST LICENSEE PROJECTS27bravura marketing director,13 had assiduously begun to cultivate mobilephone manufacturers, Nokia included. Psion Software was certainly nottheir only choice of partner for collaboration at the time (just as Symbianis by no means the only choice today). However, the company wasperfectly positioned, with just the right product at just the right time in theevolution of the mobile phone market. It has succeeded remarkably wellin extending that early lead into a commanding position in the market.2.6 Background to the First Licensee ProjectsThe first Organiser shipped in 1984. Over more than ten years, Psionhoned its hardware and software skills and learned through three completeiterations (Organiser, Organiser II and Series 3) what it took to create acomplete software system for mobile, battery-powered, small-footprint,ROM-based systems, before embarking on the 32-bit EPOC operatingsystem from scratch.
The Series 5 shipped in June 1997. Almost exactly ayear before, Psion’s software division had been spun off into a separatecompany, Psion Software. Almost exactly a year afterwards, in June 1998,Symbian was created as a joint venture aimed at bringing EPOC as a newoperating system to mobile phones.14Even before the Series 5 project completed, licensees of SymbianOS from at least three companies were waiting in the wings. There aredifferent versions of the story, but they all agree on the main points.Martin Budden:As I heard the story, Nokia were in the market for a new operating systemfor their Communicator and they approached us. I know that Juha wasinstrumental in brokering the deal, but it was Nokia’s idea and I rememberthere was a time when we were told Nokia were coming to see us.
It wasn’texactly ‘smarten up the office’, but you know, ‘if they ask questions, give goodanswers’. It was Nokia that was strongly in favor of bringing in other phonemanufacturers to form a consortium, or that’s what I understand. They fairlyquickly brought Ericsson on board and then Motorola got on board at the lastminute, and that was also quite significant.13The legend within the company when I joined in 1997 featured Juha going off to coldNorthern climes, sharing saunas and vodka with Important People and coming home witha Nokia deal in his pocket. Juha was later tempted away to Microsoft to lead the WindowsSmartphone effort.14[Tasker 2000, Chapter 1] provides a definitive history.28THE HISTORY AND PREHISTORY OF SYMBIAN OS‘Symbian Day’ was June 24th 1998 and the Psion share price rocketedon the news (causing much excitement in the office over the next fewdays).
A few days before, we had delivered the first free SDK for whatwas still Release 2 of EPOC.15 Version 5 was still a whole year away andthe first Unicode release, ER5u as it was known, was a step further still.16This, arguably, was Symbian OS ‘version 5’ (although it was never calledthat), the first operating system release that was ‘fit for phones’, althougheven then the first phones were still a year away from market. The firstdesignated Symbian OS release, v6, appeared in Spring 2001.From a public perspective much was made of the Nokia versusMicrosoft angle and some commentary viewed the creation of Symbianas an attempt by Nokia to build an alliance against Microsoft. But it seemsjust as meaningful (and more useful) to view it more as a case of Nokiamaking a shrewd move to work with competitors to consolidate and growa new market (that of highly capable, multi-function, phone-enabledterminals: ‘smartphones’ in other words), at the same time enablingNokia to focus on what it clearly saw as its strength, the user interface.The evidence [Lindholm 2003] is that Nokia viewed the user interface asthe critical software design factor for the phone market – if not the keydeterminer of success then at least a critical one – and also that it viewedthe user interface as its critical strength.However, even before the Nokia approach, Psion had been activelyevolving its own strategy and there is no doubt that a fundamental shiftoccurred after the start of the Protea project, leading to the spinning offof the software division to open the way for software licensing.
The teamled by Howard Price moved across quite late to the EPOC project to startwhat turned out to be the last rewrite of OPL, this time for the 32-bitplatform. By then the company’s focus had shifted quite noticeably.Howard Price:The big thing in every team meeting was, ‘Where are we with licensees?’ Sowe’d go to the senior team brief and a lot of the talk would be about winninganother licensee, or that the licensees were getting unhappy because of thisdelay or that delay, or that they were worried we were delaying their productsto concentrate on Psion work – the Series 5 project was running worryingly15Giving SDKs away to developers was still considered controversial within the companyat that time.16ER5u was an interesting experience to live through, a complete rebuild of a systemwhich still, at that time, did not routinely build from source (as it does these days, withnightly builds of multiple variants from a single master codeline), with the ‘wide’ flag set forall components in the system so that all descriptors, text data and resource strings (anythingwith text, in other words) built ‘wide’ using multibyte (UTF-8) Unicode text encoding.
Acomplicated system of ‘baton passing’ was evolved to follow the dependency graph upthrough the system and ensure that for every component, all dependencies built first; nottrivial in a system which still harbored some awkward circular dependencies.BACKGROUND TO THE FIRST LICENSEE PROJECTS29late. It took a year longer than planned for us to ship and the licensees werewaiting.The licensee strategy was squarely pitched at the phone market.Andrew Thoelke:The Series 3 family had been Intel based, but even at that point in ’94, ’95the view was to migrate towards mobile and cellular applications.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
