These are the old pages from the weblog as they were published at Cornell. Visit for up-to-date entries.

August 13, 2003

In Defense of DARPA

Joi Ito points to the resignation letter of Pointdexter, who headed the Total Information Awareness Office of DARPA. DARPA suddenly has become a household name, after the FutureMap publicity disaster. FutureMap was a research prototype to investigate the use of market driven techniques in the evaluation of Middle-East intelligence reports.

DARPA, the Defense Advanced Research Projects Agency, suddenly has become the target of politicians and the news media. DARPA was only known to a part of the bigger public as the agency that funded the birth the TCP/IP/DNS protocol suite, but in general there aren't many outside the commercial and academic research labs that can really grasp the scope of DARPA's activities. Charles Cooper recently came to the rescue of TIA/DARPA with a CNet article.

I not sure whether everyone really realizes the impact DARPA has had on computing as we know it. Not just network protocols and networking applications, but for example a signifcant portion of operating systems research and development was funded by DARPA. Unix BSD was developed with significant DARPA funding, as was the development of the Mach operating system, basically resulting in that without DARPA there would be no MacOs X, or any of the *BSD variants. Real-time Linux wouldn't be what it is today without DARPA funding. DARPA has been the incubator for hardware development as well as lots of Middleware technology. Fault-tolerant computing as we know it wouldn't exists if it wasn't for the funding DARPA provided. Same goes for a lot of security and intrusion tolerance work.

DARPAs choice for projects to fund has always been driven by the high-risk - high-payoff paradigm. They really were always looking for research that pushed the envelop significantly, and as such quite a bit of research failed, or didn't reach a larger deployment. But the systems that did pay-off did so in a big way, and many of systems have been the basis of a successful software and hardware companies. Market driven research as to be used in FutureMap is an example of far-out research, and it should be evaluated in that context.

It is not that DARPA should be kept free from criticism, many actually feel that DARPA should go back to taking more risks in the research projects, and shed some of its current bureaucratic nature. The appointment (and resignation) of Pointdexter was an example of a growing impact of politics on the nature of research.

Posted by Werner Vogels at August 13, 2003 10:14 PM


The author of the CNet article is Charles Cooper, not Alan Cooper.

Posted by: Anonymous on August 27, 2003 02:35 AM