What the FBI doesn't know can kill you. That, at least, is what we've been led to believe since 9/11.

Had agents in Minnesota been allowed to search Zacarias Moussaoui's computer, had the Phoenix office memo warning of Al Qaeda members enrolling at U.S. aviation schools filtered up the chain of command, had the internal computer databases done anything but the most rudimentary searches, maybe, just maybe, things might have gone much differently 18 months ago. But for the failure to connect those proverbial dots, 3000 lives might have been saved.

The idea that 9/11 could have been prevented heightens the tragedy, of course, but also invites all kinds of speculation: if we accept the premise, then there must exist some deliberate course of action that we should now take. Experts and officials have spent the last year and a half trying to figure out what that course should be, but have reached no clear agreement.

Understandably, much attention has focused on the Federal Bureau of Investigation (FBI), both for what it could have done but didn't, and for what it should do now. The post-9/11 revelations of mistakes and mismanagement only underscored what many had said for years: that the bureau's fundamental organizational, cultural, and technological deficiencies have bred a swarm of high-profile gaffes [see FBI Under Fire, PDF format ] and render it unsuited to the intricacies of fighting terrorism.

As for improving, under Robert S. Mueller III, sworn in as director a week before the September 2001 attacks, the bureau has announced a broad agenda of technological initiatives and long-sought organizational reforms. Some of that is merely catch-up, like upgrading the FBI's antiquated computational infrastructure to an acceptable, but by no means advanced, standard. More exploratory efforts, though, in investigative data warehousing and information-sharing networks, could, if successful, place the agency for a change on the technological cutting edge. As the nation's leading agency for domestic terrorism and federal law enforcement, Mueller has said, the FBI "should be the most technologically proficient investigative agency in the world."

First, though, he'll have to bring the bureau into the 21st century, technically as well as culturally, and beef up its capabilities in intelligence gathering and analysis. Then he'll have to fend off a growing chorus of critics intent on divesting the bureau of its domestic intelligence responsibilities. Also to be addressed are the concerns of privacy and civil liberties advocates. With the FBI's powers greatly expanded under the USA Patriot Act, they fear the innovations being put into place are just the first steps in setting up a police state.

Lastly, there are the many unresolved technical questions: is it really possible to build a system that can precisely identify a crime's precursors, when the would-be perpetrators are doing their utmost to be untraceable and unpredictable? And is the FBI the right outfit to build such a system?