On Dec. 8, 1941, the Signals Intelligence elements of the U.S. government were in a quandary. How could Pearl Harbor have occurred, given their spectacular success in breaking Japan's top diplomatic code? The answer was simple: In their zeal to succeed in the diplomatic arena, they had failed to place enough emphasis on Japanese naval traffic, the one source that could have given some warning of the attack.

Although they are more than a half-century old, Pearl Harbor's lessons are relevant today. Indeed, given the current debate about the role of intelligence in assessing Iraq's efforts to develop weapons of mass destruction and in disasters such as that of Sept. 11, 2001, history compels us to ask if we are once again allowing technological success to blind us to an emerging threat.

Consider today's "diplomatic code" analogue: voice communications. The numbers are staggering. In 2002 there were about 180 billion minutes of international phone calls representing the communications of roughly 2.8 billion cellular and 1.2 billion fixed-line subscribers worldwide. Excluding North America, the bulk of these communications are made possible by several hundred satellites orbiting the globe.

Despite these daunting statistics, our nation has turned the challenges of accessing and "mining" these communications to discern the intentions of our adversaries into an immeasurable success. But ironically, as with the focus of code breakers before Pearl Harbor, this very success represents the single greatest threat to the continuing viability of Signals Intelligence. Specifically, it inhibits critical investments to create equal capabilities for the communications medium of choice for a growing number of commercial providers and their customers: the Internet. The reasons are threefold.

First, the demand by Congress for better business practices in the acquisition and development of systems inevitably forces decision makers to choose investments that provide the surest near-term returns. Developing solutions to access and process communications in the high-speed world of the Internet is expensive, risky and time-consuming.

Second, customers of intelligence, such as the CIA, FBI, State and Defense departments, and the White House, have come to rely on -- and demand -- an uninterrupted stream of intelligence. But the sources and methods used to obtain this information are kept tightly guarded for fear that inadvertent disclosure will tip off our adversaries and deny us critical intelligence in the future. This means we are effectively keeping these customers in the dark about strategies for new types of collection they might wish to influence.

Third, recent perceived intelligence failures have dramatically shifted longer-term strategies toward temporary fixes of dubious value.

To gain an appreciation of the emerging challenge, consider these facts. A single strand of fiber-optic cable exceeds the capacity of all the telecommunications satellites orbiting the globe. This year alone, e-mail volume is expected to be the equivalent of 40 copies of the fully digitized holdings of the Library of Congress.

Another important medium, instant messaging, is now estimated to generate 530 billion messages per day. Complicating matters further, phone calls can now be sent directly over the Internet using a technology called Voice-over-Internet Protocol (VoIP). AT&T, Teleglobe, British Telecom, Time Warner Cable, Telecom Italia and Deutsche Telekom began offering VoIP services this year, signaling the beginning of a major conversion away from traditional telephone networks. Indeed, by 2005 enough fiber-optic cable will have been laid to carry, by similar analogy, one Library of Congress every 14.4 seconds.

What has been the reaction to this challenge? In the National Defense Authorization Act for fiscal 2004, Congress revoked the National Security Agency's "milestone decision authority," effectively stripping the NSA of its ability to directly manage its own modernization program. This authority now rests with the undersecretary of defense for acquisition, technology and logistics -- a situation that will severely restrict the NSA's ability to invest broadly in unproven but highly promising, and in some cases competing, technologies. The additional delay this introduces in acquiring such systems will be devastating, particularly since the turnover of technologies on the Internet typically spans months, not years. Moreover, the people who know the most about what is needed will be further removed from the point of decision.

The emergence of transnational and terrorist targets as national priorities has made intelligence consumers in the government eager for intelligence on demand. Tactical requirements for near real-time intelligence supporting operations by ground troops and law enforcement continue to grow. Given a dollar to spend on intelligence, the most profitable near-term return continues to come from existing systems.

This inverted cycle of demand and supply comes at the expense of laying the groundwork for future capabilities, an expense largely hidden from customers who might wisely choose to demand less from the intelligence community in the near term, given the long-term stakes.

Perhaps the most profound, and discouraging, event for the intelligence community in the past three years is the message sent by the various Sept. 11 investigative commissions: The failure to share data was the single greatest factor in the failure to detect the attack plans, a conclusion that clearly underlies the more provocative fact that no "smoking gun" was found. Indeed, the provocation is this: It would have been worse to have collected clear and compelling intelligence before Sept. 11 and not to have analyzed it than to have had no such information in the first place. And the subtle consequence is this: Analysts are hesitating to pursue new sources for fear that they will be unable to process the resulting data without new tools and technology -- investments that are difficult to justify because their return is unproven.

In his influential book "The Innovator's Dilemma," Clayton M. Christensen describes how extremely well-managed, successful companies can ultimately fail when confronted by "disruptive innovations." Such innovations are often mistaken by incumbents as too expensive, without proven applicability and lacking a sufficient customer base. Much like the code breakers shortly before Pearl Harbor, we risk not heeding the disruptive shift that is occurring with the growing dominance of the Internet. History -- and best business practices -- obligate us to encourage "disruptive innovators" to create new sources and methods for intelligence and let the natural evolution of innovation deliver reduced costs, greater capability and larger markets. To succeed we must demand far less near-term intelligence product from the Signals Intelligence community, give it control of its resources and allow it to plan for a disruptive future, a future that is presaged by videos that show an Afghan warlord exhorting his terrorist followers not to use satellite phones for fear of American capture.

We would do well to understand that today's cryptologists must no longer look only to the sky for answers.

The writer was a cryptologic mathematician with the National Security Agency for 21 years, the last three as the NSA's senior technical director. He is with RABA Technologies LLC, an information technology consulting company in Columbia.