Interview with Peter Horne

Horne is Managing Director of the British company Apricot, which he joined as Technical Director to set up a research and development div- ision in February, 1983, when Apricot was the number one British computer manufacturer. Founded in 1965, like Acorn Apricot lost an early lead - it continued with a proprietary system long after most of the world had gone to PC-compatibility. It was bought by Mitsubishi in April, 1990.

Horne got his degree in electrical and electronic engineering, and has a PhD in statistical pattern recognition. He was heavily involved in early networking at CERN (1975-78) and spent two years at Cambridge Scientific Instruments designing a multiple-processor system to analyse imagesirom chess X-rays and satellitephotographs. Horne became Managing Director in 1989.

Peter Horne must be one of the only directors of a computer company who would start an interview by admitting that computers are a disappointment. "The fact is that the usefulness of computing has not been as great as everybody thought it should be," he says. "If you measure it in terms of the profitability of companies, then it's very difficult to say it's actually achieved anything."

Horne sees problems with today's networks - and says the industry isn't really equipped to handle them. This, he says, has become apparent as more and more large corporations want to downsize from mainframes and minis to PC networks. The reason, he thinks, is that traditionally all the complexities of wiring things together were left to the major manufacturers, like IBM, DEC and JCL, who "basically did most of the integration work for them."

What the smaller organizations did was "bespoke software and project management, but their knowledge and depth of knowledge on overall systems integration - middleware software and how to make it work for an organization - I think is fairly poor."

The answers needed are different now. "When you come to our world, it isn't go out and design another bespoke program to solve the problem, it's saying, somewhere in the world is a solution to this problem, where do I find it? This is a big cultural change."

Moving into consulting, he says, "is everybody's answer, to some extent. Even Apricot is saying we're going to have to do a lot more than we thought we had to."

Horne was just discovering this sort of problem when in 1982 he saw what he describes even now as a "strange job ad." A company in Birmingham was looking for a chief engineer; what was unusual was that the remuneration included shares in the company.

"I already knew of the company - it was the only company likely in Birmingham - because I had bought a product called the Sirius from them for my own business."

At the time Apricot was the number one computer manufacturer in the UK. IBM hadn't started selling here yet, and Apricot's computers were Sirius compatible. Who knew? Even in the US the leaders were Tandy and Apple; Apricot, then known as ACT, discussed with Tandy the possibility of launching an Apple-beater in the US. Tandy declined to get involved, and Apricot made the attempt by itself, unsuccessfully.

"The US market had gone compatible by then," says Home, "so we didn't sell many products." In 1984 Apricot, by then renamed for its computers, introduced early 286s - but even these were still Sirius compatible.

"By then we knew it was too late," says Horne frankly. "Those should not have been Sirius compatible, they should have been IBM compatible, so we were a year late." It was June 1986 before Apricot started selling IBM-compatible PCs.

"We were pretty successful with the product despite that, but many of our larger resellers had by that time shifted away, and were carrying Compaq and IBM, so we'd lost a lot of our channel that we should have maintained. We ended up with a very interesting channel of resellers in the UK, very competitive, very VAR-like, compared with the one that had gone the box-shifting route with the PC."

Compatibility remained an issue at Apricot even then, "like a smoker giving up smoking." The company was not just going to be compatible: it was going to be really compatible. And that's how it came to decide to make MCA machines instead of ISA. Horne believes Apricot is almost the last company making its own PS/2 clones.

"We waited too long on not becoming IBM compatible," Horne sums it up, "and when you've made a mistake once you try not to make it again. But the whole industry flipped. Today we take a pragmatic approach; we listen more to customers, like everybody else."

Selling MCA clones opened the way into some of IBM's accounts, however: "Like Compaq in 1984-1985, we were the only company offering complete compatibility." There is a common impression that PS/2 architecture is somehow proprietary, or at least more expensive to license. Horne says this is absolutely untrue.

"There is so much bunkum and bullshit about licensing," he says. "There is no such thing as licensing MCA. IBM has a straightforward patent licensing policy that covers three different major areas and categories. One category covers about 9000 patents, and you take out one licence and it covers everything. Anybody can take it out; it costs maybe $25000; it's not expensive. It's no more expensive to make PS/2 clones than PC clones."

Other than the licensing, though, PCs are cheaper, if only because there are more parts available from more suppliers, which tends to drive the price of parts down. Apricot now makes a line of clones, too. "For the desktop user, neither MCA nor EISA has any benefit for the customer that they can clearly see today."

For Home, the argument over which architecture to use is as unimportant as the arguments he used to listen to ten years ago between Ethernet and Token Ring.
The exception is servers.
"In the server you do need a decent bus, because you'll be connecting a lot of high-performance peripherals to it. But, "Most customers make their decision based on price, performance and functionality, as opposed to what the little piece inside the engine does."

The way Horne tells it, Apricot has something in common with Apple: it tries to put as much as possible in the machine so all the customer has to do is plug the thing in. The company started building in networking in 1984, and its recent multimedia machines come complete with built-in sound. But trying to get customers to see the benefits of that is uphill work in a market which has been trained to consider price first, second and third.

"The bad news of the compatibility market is that innovation is almost out. As a small company, it's very difficult innovation because the customer says the standard machine looks like this, and it has eight slots, and unless I can plug cards in, it's not a PC, is it?" So small companies have to conform, and the only hope for change is if the larger manufacturers do the innovating. Unfortunately, IBM and Compaq, who could lead the way to change, are in fact following the trend by launching low-cost lines.

"We feel we're a lone voice in the world saying there's a better way of doing this, because the world at large doesn't want to listen. The only thing it can think about is price, and therefore innovation is getting pushed even further away. Products are being driven down in all aspects of quality for the customer. Clearly there's a good side - the customer can now buy great technology at a very low price - but there are penalties to pay for that in the future."

Horne believes having Mitsubishi as a parent helps Apricot. "We can afford the investments and the time to start the education, but it's a tough battle. I do see that the next couple of years are going to be really traumatic for the industry, but unless it does change, in two years' time a lot of customers are going to be really pissed offwith the industry." Horne has problems with an even more basic tenet of IT marketing: the idea that people throw out computers every couple of years and buy new ones.

"You go in to a corporate customer and see computers that are eight or nine years old. Nobody throws computers away. None of them upgrade them, either, and they won't, because you can't upgrade part of a PC if you're going to use the whole thing." When people talk to him about upgrading every two years, he says, he asks them, "Look at that chair. Do you throw it away every two years?"

"You don't do that in an organization; you move it to a suitable function." He himself has run a 386 at 25 MHz for the last four years, even under Windows.

It's fast enough," he says, "because I can't type faster than I can think. We're creating a demand and expectation and a worry about the machine being superseded the day they buy it that is unreal." The obsession with speed carries over into the way the industry produces software, too.

"If only by now we could have accepted a level of - not wastage, but inefficiency.. With the power we have now and the amount of memory, do we really have to rewrite every service routine and interrupt routine every time we change the operating system?" Several organizations have goals to produce libraries of standard routines, but, he says, "It's still a long way off.

"Every time, someone says, 'but if I change a byte there I can make it go so much faster,' so there's a tendency always to improve it. But that introduces the next problem. So somehow we need the wheel, the reusable part, the lyre again, that doesn't change - maybe minor things change, but basically you're not going to get it dramatically wrong. Like a 747 - that's been going for 20 years with only minor refinements, and at least you've got all that history of reliability behind you, whereas software basically gets rewritten every time."

We need reassessment: "What we've done in many cases, especially in the PC market, is make word processing prettier. But are the words any better? Or the quality of the information coming out?"