Tony Durham meets the men directing Microsoft's innovative research programme in Cambridge
The retired head of the Cambridge University computer laboratory holds forth as he paces the room, while the head of Oxford's equally prestigious programming research group, also recently retired, makes pertinent interjections from his chair. You can probably imagine the scene: oak panelling, sherry, footsteps echoing across the quad. And you would be completely wrong.
We are not on university premises but in the spacious modern offices of Microsoft Research, where the main corridor seems implausibly longer than any of the surrounding Cambridge streetlets. The city centre site is not perfect, lab director Roger Needham insists. The nightclub downstairs prevents staff working late into the evenings. There is no car parking, so many researchers walk to work. Admittedly, on Microsoft salaries they can generally afford city-centre homes. Working here begins to sound increasingly inviting.
Needham was hired by Microsoft in 1997 and has assembled a team of 48 researchers, many of them former academics. Tony Hoare joined the lab in October 1999 after a distinguished Oxford career, just weeks before he gained a knighthood in the new year honours list. He is one of the elite band of 40-odd computing researchers to have won the Association of Computing Machinery's Turing Award.
Microsoft told Needham to hire the best people and have them do what they are good at. A glance through the personal homepages on the lab's website suggests he carried out his brief well.
Needham's closest colleagues are in Redmond, Microsoft's home in Washington state, San Francisco, and Beijing. Each lab is different. Cambridge has nine European nationalities and many mid-career researchers. Beijing has an all-Chinese staff, mostly young postdocs on short-term contracts. The insecurity of the job does not deter them. "It is a good thing to have on your resume, like being an assistant professor at Massachusetts Institute of Technology and not getting tenure," Needham explains.
"Our job is to develop technologies that may turn into products in five to ten years' time," he says. He also believes it is the lab's duty to help out with more immediate projects when it can, but is stuck for quotable examples. "There is one extremely beautiful one that we are not able to talk about because the object in question has not been announced." But he is clearly pleased that Redmond is not only listening but beginning to see a return on its investment in Cambridge.
Last week a Microsoft Cambridge re-searcher, Jonathan Hardwick, announced a result with important implications for information security and electronic commerce. Cambridge and Redmond researchers have found a way of factorising large numbers, and thereby potentially cracking the widely used RSA encryption system, using a cluster of PCs. Previous methods needed a supercomputer, or an awful lot of time, for at least one stage of the process. A job that needed an elephant can now be done by ants. Needham warns that it may now be time for financial companies to start encrypting their data with bigger numbers than the 512-bit keys they currently use.
The existence of this lab and its success in luring academics could be a sign that universities are no longer such a wonderful place to do computing research. The days are gone when computing researchers could try out a new network protocol or operating system on willing colleagues campus-wide. "Users in universities are very conservative," Needham says. "They do not want to be experimented on."
So what is left for university researchers to do? "Theory," Hoare ventures, while Needham rolls off a list of applications, such as speech recognition and security, where Cambridge (and no doubt other universities) are making important contributions to practice.
Needham believes that practice stakes out new territory before theory consolidates the gains. Many branches of computing, such as high-level programming languages, have begun with a technical tour de force. When John Backus invented Fortran, the first scientific programming language, he had no theory to work by.
After that came a "black art" phase where a number of people were quite good at writing compilers (which translate a programming language into the machine's own code) but they could not explain how they did it. Gradually the subject matured into an apprenticeable skill that teachers could transfer, though they still did not really understand it.
Only then did the theory of programming languages develop. "It is difficult for the theorist to devise theories of something that does not yet exist," Needham argues. Theorems were proved, textbooks were written, and mysteries dissolved. As Hoare says, "the sort of compilers we were writing in the 1960s have become a term project for an undergraduate".
Needham believes that university research develops valuable attitudes and skills even when it fails to yield bankable intellectual property. "The most important thing coming out of university research is the people - the PhD students, the experienced postdocs," he says.
Hoare sees another vital, continuing role for universities: they must teach people to be discontented. "The only hope Microsoft has of surviving is that people are discontented with last year's release." www.research.microsoft.com/cambridge