An unquiet Frenchman and his small software sidekick

Wendy Grossman meets Philippe Kahn, the founder of Borland, who now aims to sell software over the Net

In an industry where chief executives are either scrawny nerds with glasses or men in suits who make the most exciting technology sound like grey flannel, Philippe Kahn has always been an exception. He plays jazz, he flies stunt planes upside-down, and he arrived in the software businesses as a French mathematician.

Another difference is that he's prepared to admit his mistakes. Until early 1995, he was chief executive of Borland, for a long time the third of the "big three" PC software companies (behind Microsoft and Lotus). But Borland lost first its financial stability, then its position to Novell (which bought two of its leading products) and finally its chief executive. The mistake that preceded that departure - the takeover of Ashton-Tate - was "a big lesson", according to Mr Kahn.

Now, travelling with a Selmer sax he likes to practise on in his hotel, he is in town to promote Starfish Software, a company he co-founded last year. He looks eight years younger and a generation happier. Ask him if he misses Borland, and he giggles and says, "Not at all."

Mr Kahn never set out to run a company; he never even set out to work with computers. He was born and raised in Paris, and studied maths in Zurich, where he was a student of mathematician and computer scientist Niklaus Wirth, at the same time that the first compiler was designed for the now well-known programming language Pascal. He was teaching maths in Grenoble at the right time and place to come across the Micral, the first ready-assembled microcomputer. Very quickly, he realised that the future lay in programming and California's Silicon Valley. So, in 1982, he headed there, lacking even a Green Card, hoping to work with his heroes: the technicians at Apple.

Instead, he wound up founding Borland, which began with a magazine ad offering a microcomputer-based Pascal compiler after other firms - including Microsoft, which was big even then - declined to license and market it for him.

Borland was unconventional, like Kahn himself. Its main product, Turbo Pascal, cost $50 when most software cost over $400. In 1984 it released Sidekick, the first "productivity application", designed to replace the calculators, appointment books, phone books and notebooks that clutter most desks. It was Borland's third product, and the one that made it famous. But there were grumblings among Borland executives about Kahn's predilection for jazz jam sessions when they thought he should have been poring over printouts.

Sidekick for Windows, which Mr Kahn was heavily involved in designing, arrived in 1994. The recently released version 2.0 is the first new product for Starfish, whose mission in life is to produce software that fits on a single floppy disk and is simple enough for anyone to use. Its other product is Dashboard, which is intended to tidy up computer screens and tell users how efficiently their computers are running.

"It's about craftsmanship," says Mr Kahn, to explain the logic of products that fit on one floppy disk - up to 1.4 megabytes - when so many occupy hundreds of megabytes on CD-Rom. "Also, if your applications programs are small then you can distribute them over the Internet with reasonable download times." The value of this becomes clear during a demonstration, when he finds that a virus on his computer has infected a test version of Dashboard. He immediately makes a phone call to wake someone in California - at about 4am their time - to demand that a fresh copy be sent via e- mail at once.

"We sell intellectual property," he continues, the task done. "We don't need factories. There's just 45 people working on making the best-quality software they can. It takes about 20 minutes to download the product over the Net at 14.4 kilobits per second. Most of the copies we've sold so far have gone through standard retail channels; but as that grows, so does the business over the Net."

Intellectual property became his cause celebre. In 1991 Borland and Lotus fell into dispute over a feature in Borland's spreadsheet, Quattro Pro (since sold to Novell) that interprets macros written for Lotus 1- 2-3, the program that made both Lotus and the PC successes.

Lotus claims that this feature contravenes its product's copyrighted "look and feel". Mr Kahn contends that compatibility and "interoperability" of software is vital for users, and that Lotus is attempting to stretch copyright to cover functionality. That, he says, is equivalent to copyrighting the chord changes a musician uses in playing a song. "If chord changes were copyrightable," he asked industry magazine Microtimes, "when's the next time you'd write a new song?"

Lotus won the first round of litigation. In March Borland won its appeal. Lotus now intends to take the matter to the US Supreme Court, but it's widely believed that the court will refuse to hear the case. Kahn thinks that - principles apart - enough money has gone to the lawyers. "It's yet more money spent on litigation rather than innovation."