No one has nominated an ideal research lab. It seems we all must "work" for a living, except for isolated individuals with research grants -- the Presidential Young Investigator (PYI) awards, for instance.

Many labs started with good intentions -- freedom to choose projects for maximum scientific impact -- but convert to applied or short-term research within a few years. Many AI labs were founded to draw recent AI graduates and clone their LISP expertise, but managers planned from the beginning to extract some value within two to three years. Semi-independent labs -- ICSI, RIACS, CSLI, MCC -- often serve philanthropic, tax-sheltering, or public-relations functions when new, but must look for new sugar daddies when good intentions fail to produce a sufficient stream of results. (I'm told that RIACS is down to 1/3 of its former staff. Founder Peter J. Denning has moved to George Mason University, and Dorothy Denning has left DEC WRL/SRC for Georgetown University.)

Physics and astronomy labs draw funding for "pork" and national prestige, plus defense and commercial potential. Pure research labs in computer science survive mainly in hardware domains, where patents protect the parent company from "giving away the store" in scientific publications. (If it weren't for the patents, companies would do just as well to let universities or start-ups pioneer new technology.) Now Microsoft and Apple are scrambling for software patents, and Microsoft has coincidentally opened a research lab. Hmmm. Perhaps Apple could be persuaded to do the same.

I was wrong about Xerox PARC (Palo Alto, CA) having a hiring freeze. Mark Weiser (mark_weiser.parc@xerox.com) is looking for researchers in systems work: hardware, languages, operating systems, and networks. PARC has been at steady state since 1980. It MAY be true that AI groups have not been hiring recently, but that is not an across-the-board freeze.

Mark adds that Xerox is not philanthropic, it's just willing to bet on long-term returns. The Microsoft lab may not be in the same league. Jim Mallory's Newsbytes report [Computer Currents, 9/10] quotes Nathan Myhrvold as saying "We're going to look at stuff that is two to five years out and focus a lot of effort on it." Mark doesn't consider that basic research, given that it has taken more than 3.5 years to develop Windows 3.0.

On the other hand, Microsoft is saying that there will be no management pressure to undertake projects with clear profit potential. Rick Rashid says that this will be no skunk works or Windows project. "This is an opportunity to take a fresh look at what computing can be like, how computers will be used in the next five to ten years." He's modeling the lab partly after Xerox PARC, and expects to publish in scientific journals. [Rory J. O'Connor, SJ Mercury, 9/15.]

(BTW, Mark Weiser has heard that Brian Berhad will be inheriting Rick Rashid's Mach DARPA contract, ten Ph.D. students, and staff. Brian is a recent graduate of UWashington, having worked under Dr. Lazowska on specialty operating systems.)