09 May 2006

Surviving Big Blue

In the early '80s technology manufacture seemed to be thriving here in the UK, particularly among the smaller startups that as an IT journalist I spent a lot of time talking to.

But by the time I wrote this sadly prophetic piece for "Which Computer?" in 1985 the impact of US manufacturers was already beginning to take its toll. The reference to the woes of ICL is particularly poignant for me, even today. In the late '70s I'd been researching an industrial film for ICL, and happened to mention "micros" (as PCs were called back then). I was impressed by the advent of the Apple II, and warned, in my smart-arse newbie way, that this "revolution" would prove a threat to ICL's business. How the rotund, stripe-suited mainframe salesmen chortled. "Mere toys," they said. Hmmm...

Sir Clive's QL gets a mention here. I had one to play with, but never used it seriously. But my successor as script editor of "Doctor Who", Eric Saward, bought one and I understand made good professional use of it for years after. I also reviewed the ICL spin-off from the QL, the One-per-desk. It was quite the most depressing piece of hardware ever to visit my office.

(Incidentally, the @string format headers remind me that I was probably composing this in Perfect Writer, and running it through a formatting program before printing it out for my editor.
)

PS: I see I even manage to slip in a quote from "Hamlet" here. How classy is that?

[Surviving Big Blue][wc?][25-Mar-85][chb]

@string(Today = "25 Mar 85")
@string(ShortTitle = "Survivors")
@string(ArticleTitle = "Surviving Big Blue")
@string(filename = 'BigBlue.WCH')
@string(header = "article")
@string(subpage = "")
@include(a:general.hed)

In the twenties it was wireless sets, the fifties, tape recorders, the sixties, high fidelity. For the small and not-so-small technological entrepreneur, prepared to manufacture, import or add value here in Britain, each phase has opened up a booming market ... and within a few years, slammed it shut again as the heavy international operators loomed onto the scene. The mushrooming of micro manufacturing at the turn of the decade repeated the cycle. 1982 brought IBM and its Personal Computer decisively into the game, and the casualties have been mounting ever since.

In the UK three things happened over the next few years. Companies like Keen, with a heavy commitment to outright if indifferent alternatives to the IBM philosophy found their market eaten away; others, like Almarc Data Systems, in developing and manufacturing IBM-lookalikes, diverted too much of their resources from their tried-and-tested stable product. Keen, Almarc and many others are now just memories.

The third thing that happened was that, remarkably, some UK computer companies survived.

The larger internationals -- Olivetti, Sperry, Burroughs, Ericsson et al -- that began with micros of their own design have been steered by diminishing sales returns into the production of '100% compatibles'. All of them promise, and most deliver, some technical superiority over the distinctly dull IBM product, but the design constraints limit the possibilities.

Wang is an exception: it continues to rely on a machine using the same processor family and operating system as the PC, but unable to run standard IBM software that addresses the hardware directly (as most of the Top Pop packages like Lotus do). Instead it offers innovative industrial design, and special goodies like TV image capture.

Meanwhile in another part of the forest, Apple plows its own distinct and costly furrow with the Macintosh.

Let's call these three strategies stringalong, stayclose and stayaway. The same options are open to the UK survivors, although most lack the considerable marketing muscle to stringalong, and stayclose is increasingly dangerous for both manufacturers and importers as it appears to be the ground picked by Japanese manufacturers like Sanyo for a pile-'em-high export drive.

Cifer, the mid-seventies VDU company turned micro manufacturer, long ago formulated the theory that in a world of intense competition for the low and mid micro markets, where economies of scale virtually guarantee the eventual domination of the Japanese and the Americans, there would always be room for excellence. The philosopy has made them stayaway survivors... but only just. Following heavy sackings last year at its worksite in Melksham, Wiltshire, Cifer, has now made further cutbacks of a third of its workforce, accompanied by major managerial reorganisation.

What went wrong? Physically their Series 9000 is a highly tuned, superbly designed computer that draws inspiration from the Hewlett-Packard range. But the company put all its eggs in the single basket of Unix, the operating system that is tipped (at least by AT&T, its proprietors) to take the micro world by storm once sufficiently advanced micro-processors are available.

However Unix has yet to prove itself a commercial success. Critics argue variously that it is far too advanced for business users, or alternatively, too old and clumsy to cope with the sort of pictorial interfaces customers are rapidly coming to expect.

For its often fanatical adherents in the academic and scientific worlds, Unix is a philosophical adventure, and a sublime environment for the development of software. Even they would concede it takes time and trouble to install on new microcomputer hardware, and when up and running steals so much of the processor's power for its own internal ruminations that some have dubbed it 'the octopus in your tank'. While Cifer convalesces from its wrestling match with the octopus, doubts are bound to deepen as to whether the world is ready for Unix, or Unix for the world.

Cifer's level of commitment to Unix may be a misreading of the undoubted market interest in the product. "I get asked for Unix a lot," says Trevor Crotch-Harvey, Marketing Manager of TDI, another company involved in 68000-based machines. "A few customers need it specifically -- the other ninety percent only need the assurance of its availability. Unix becomes a sort of general indicator that the our Pinnacle and Stride machines aren't going to be left behind".

For that reason TDI are currently porting Unix System V, although few of their customers show signs of using it "They just need to know you've got it on your box. We don't expect to sell too many Unix machines -- but we expect to sell more machines because we've got Unix!"

TDI is an exemplary exponent of the stayaway strategy. The company began life as a time-sharing bureau, an activity that now accounts for only 5% of its business. Hardware and software distribution in equal parts make up the rest -- the Pinnacle and Stride micros on the one hand, and systems software and development tools on the other.

In 1982 the Stride, then known as the Sage, was the first 68000 multi-user micro introduced into this country. TDI are sole UK distributors for these Nevada-made micros, but the company's relationship with the more powerful Pinnacle, another 68000 machine, is closer still, with TDI sharing ownership. So far manufacturing is confined to Dallas, Texas, but most of the systems software was developed here in the UK.

Crotch-Harvey doesn't remember ever making a conscious decision to stay away from Big Blue. "When the IBM PC came out we already had a machine that was miles ahead of it technically. You might call it naivety, but it never occurred to us to take the backward step of offering anything similar."

This isn't just marketing hype. The central processor that the TDI machines share with Cifer's Series 9000, the Motorola 68000, dubbed 'the programmer's pet' is a high powered, true 16-bit engine, with 32-bit internal resources far beyond the 8088 that drives the IBM PC. But, as IBM knew all along, and others have since learnt, raw power means nothing to the mainsteam end user, who looks for a stable and rapidly expanding software base.

How does TDI stay away from IBM without staying away from customers? "The most important thing is to have some major compensation for incompatibility," says Crotch-Harvey. Micros like the Pinnacle allow users to run heavy number-crunching programs hitherto unthinkable outside a mainframe. Porting such applications is often surprisingly simple -- the source code is usually in Fortran and all that is needed is an efficient Fortran compiler on the target machine.

The result is a very price-competitive alternative in a number of specialised applications, like the simulation of complex real-life situations. One example increasingly used by designers and architects before the first brick is laid, is analysis of the flow of customers and goods around the checkout counters of a supermarket.

Since the early seventies there have been mainframe packages that animate the effects of random peaking and bottlenecks, but the massive repetitious calculations they require run far too slowly on the standard office micro, and threaten to seize too much of the resources of shared mainframes. Porting them to high-powered micros like the Pinnacle that may become dedicated simulation machines can be the ideal solution. And often at something like a twentieth of the traditional cost.

TDI's micros also have a part to play in the corporate office, where they can be found running the BOSS business-oriented operating system as fast multi-user machines. The initial cost of the central processing unit is a little under £7,000, but adding users is simply a matter of plugging in additional terminals. A typical six user system will run to around £11,000, less than £2,000 per user, with all the benefits of shared resources like hard disks, printers, and of course communal data.

Local government particularly offers scope for suppliers of non-IBM multi-user micros. John MacGivern is Sales Director of Casu, a stayaway company specialising in this field. "We keep well away from compatibility with the PC and specialise multi-user systems. For most of our customers these are cheaper than stand-alone micros. Typically they write their own applications, so not being able to run ready-made packages isn't a problem. In any case, operating systems like Concurrent CP/M and Boss are a far richer environment for developing software".

HM Systems, a small London company that operates under the broad umbrella of the Grand Metropolitan Group, believes that multi-user systems should be multi-processor too. Their Minstrel micro differs from a normal multi-user system by dedicating a separate processor, with associated memory, to each user. In effect the Minstrel is a network of separate micros in a single chassis with the connection made down a fast and physically short parallel bus system housed within the machine.

This arrangement appears further from the IBM PC than any of the stayaways so far described, but for existing IBM PC users there's a have@|-your@|-cake@|-and@|-eat@|-it angle. "We don't ignore IBM," says John Kelly of HM Systems. "Our operating system, TurboDos, is compatible with a lot of IBM gear. In fact we have a facility called TurboDos PC which links any MSDOS or PC-DOS machine into the Minstrel network."

This means that the Minstrel can operate as a fast file server for IBM PC's, although Kelly prefers to put the emphasis firmly on his own machine. "In a network like that you've got a real computer for doing multiple access accounts work and archiving, with your dedicated baby IBM-style workstation attached to handle executive toys like Symphony and Lotus 1-2-3."

A stayaway approach in price as well as technology is exemplified by Sir Clive Sinclair, who used expertise acquired in bringing the toy-sized Spectrum to market to produce the cheapest possible 'serious' computer. Built around the 68008, a close relative of the 68000 that cuts costs by working with slower cheaper components, the QL offers a complete system for an incredibly cheap #600, including colour monitor and bundled business software.

Unfortunately 'incredible' became the key word during a premature release of the product. Its innovative storage system, a pair of 100K continuous loop tape cartridge called 'microdrives', became notorious for their unreliability. Since then a group venture with ICL has brought twenty years of mainframe manufacturing experience to bear on the problem. ICL's high speed photography of the tape loops in action suggested modifications that have now produced a cheap, reliable machine of real utility.

A serious threat to IBM's continued expansion in the business market? If the product itself were the only factor, the answer would be a defiant yes. Except that initial mis-marketing and an appallingly bad press have taken their toll in sales. More recently, rumours of a serious attack of cold feet on the part of Sinclair's private shareholders threaten the QL's future, and casts doubts on Sir Clive's next venture, a 'no compromise' lap-top portable.

Meanwhile ICL, brought by recent troubles into the ownership of STC, continue to push forward against IBM's domination of the PC market on two fronts. The group venture with Sinclair has sired the One-Per-Desk, a re-engineered QL with a phone. Bred to be the middle manager's Shetland workhorse, combining the communication skills of ICL and STC with Sinclair's cost-engineering expertise, the OPD has so far turned out to be the proverbial camel designed by a committee -- and a committee that seems not to have met very often. Not least among its shortcomings is an internal communications problem that prohibits text transmitted and received down the telephone line from being edited by the built-in word processor.

ICL's up-market thrust, the ICL-PC, is a classic stayclose product, but their choice of central processing chip opens up a new possibility. On the cards is nothing less than a stringalong-stayclose-stayaway machine that may solve everybody's problems -- even ICL's.

Before the arrival of the 80286, manufacturers who wanted compatibility with the IBM PC had to copy its physical architecture. The fly in the ointment is Lotus 1-2-3 and programs like it that by-pass the operating system in order to achieve more speed and to give copy protection. In the colourful jargon of the systems developers, such programs 'hit the hardware directly'. If the hardware is not as the program expects, it will crash.

Bill Gates, youthful mastermind behind MSDOS, the operating system common to IBM-PC's, stringalongs and staycloses, calls programs that include such extra-curricular activities 'badly-behaved'. This is the factor that makes an MSDOS machine like ACT's Apricot unable to run the standard versions of IBM-based software.

The 80286 can do everything that the 8088 of the IBM PC can do and more. In particular it offers the opportunity to catch these direct hardware hits and allow them to be specially treated by the operating system -- in effect mimicking the IBM PC hardware by software emulation.

Paul Bailey of Digital Research keenly promotes this approach; it is his company that is supplying ICL and others with Concurrent Dos-286. This chameleon operating system allows ICL and other 80286 manufacturers to build machines that will be able to cope with all the existing body of IBM-PC software -- and at the same time promise multi-tasking, windowing and true concurrency, three features that together allow a single micro to do several things at once.

ACT is another stayclose company that may be hoping to outflank IBM in the same way. Their current Apricot and F1 ranges are built around the 8086, an 8088-compatible chip that gives more speed at a higher cost. But this and other technical advantages, like the use of the latest 3-1/2 inch disk drives, are only too easily seen in the marketplace as purposes mistook, fallen on the inventors heads. One new IBM-PC package arrives on the market every day, and purchasers of the classy ACT machines quickly become conscious of the tantalising proximity of a growing volume of software they can't run.

The fact of not being able to put a standard IBM-PC 5-1/4 inch disk into an Apricot slot is estimated already to have cost ACT millions of pounds in software conversion fees. More is involved than copying the programs to a compatible disk of course, because intricate modifications have to made to the ill-behaved sections of code.

Despite this unwelcome overhead, ACT's flamboyant marketing, imaginative management and technically superior products have put it on equal footing with the IBM in the UK, giving each 30% of the market. The balance, though, is perilous: while IBM PCs have largely found homes in corporate offices whose occupants will remain brand-loyal into the next generation of products and beyond, ACT's distribution has sought out the nooks and crannies of the one and two man businesses, like solicitors and estate agents offices where the cost of sales are high, and customers are easily wooed away by price-cutting competitors.

Although no firm announcements have been made, it seems certain that by the end of the year ACT will be offering a Concurrent CP/M 286 machine to fuel its drive into the corporate market. But by this time it will have to contend with IBM's own 80286 office micro, the PC/AT. Announced in the States last year, the arrival over here of the AT has been delayed by chip shortages and what IBM-watchers suggest may be a wrangle between IBM UK and the American mother-ship, but nobody doubts it will soon be tangible presence.

IBM also have a software product in the offing that could pull the rug out from under the Concurrent contenders. Some 180K in size, TopView is a program for IBM PC's and AT's that beefs up the operating system to provide windowing facilities for existing 'well behaved' standard packages. Badly behaved applications have only restricted freedom under TopView, but specially written packages that take the TopView environment into account get a bonus -- a measure of integration and data exchange.

The result is something that looks very like Digital Research's concurrency. But the arrangement to share the processor is too rough and ready to support the real-time demands of, for example, a communications program running in the background.

TopView has been available in the States since the US launch of the PC/AT in August of last year. But when the machine arrived in the UK a month later, TopView failed to accompany it across the Atlantic. IBM UK won't say why, or when it will be launched over here.

Reports suggest that US customers are intrigued by the product, and may even want to buy it when IBM can explain what it's for. Digital Research are watching the muted sales of TopView with mixed feelings: if customers can't perceive the benefits of pseudo-concurrency, how do you sell them the real thing?

There are no easy conclusions to be drawn about a safe path for UK manufacturers. Optimism seems to be most justified among the smaller stayaway companies that have secured niches for themselves in specialist sectors like government and scientific establishments. At least for the moment. But the computer by definition is a general-purpose device, and as the mainstream micro technology exemplified by IBM becomes faster and more powerful, even exotic applications begin to fall into its domain.

The stayclose companies are more immediately vulnerable. If TopView and the PC/AT threaten to make waves for the UK survivors, the latest informed rumours of a brand new IBM machine could swamp the landscape. How would you like a genuine IBM lap portable with an 80286 processor, a hard disk and a 3 1/4 inch floppy (like the Apricots) for some 20% less than the current PC price?

You would? Then you'd better think what to do with those old ICL and ACT micros. And with old ICL and ACT, for that matter.

No comments: