"Garbage in, garbage out." It’s one of the oldest cliches of the computer world — and, like many cliches, it’s a cliche precisely because it is true. No matter how sophisticated the system, it is no better than the data that go into it.

It’s worth thinking about this when we think about the proposals for a national identification card. Proponents of the card like to talk about some sort of endpoint where everyone has a secure, verified, and unfakable identity card. They are a lot less interested in talking about how we are to get there from here. In fact, I’ve been unable to uncover any specifics about exactly how a National ID plan would be implemented.

So let’s perform what physicists call a "thought experiment," and imagine what would happen if the Bush Administration were gullible enough to adopt Oracle CEO Larry Ellison’s suggestion to adopt a National ID card, powered (of course) by Oracle software.

There are approximately 280 million Americans. None of them has a national identity card now. All of them will need one. How will they get them? That is, how will we insure that the identity that goes on their card is their real identity? And what will it cost to do so?

Almost certainly, we’ll look to existing forms of identification, because there’s really no other way. That means that we’ll be instructed, as we are with passport applications today, to provide a driver’s license, and a birth certificate. Those forms of identification are already insecure — that’s why we need a National ID card, remember? — so any system based on those will be inadequate.

There will be a lot of long lines, a lot of paper shuffled, a lot of computer files created and — no doubt — gotten wrong (credit reports are full of errors, and they have a financial incentive to get things right). But at the end of the day, the national identification card will be exactly as secure as a driver’s license and birth certificate, which is to say, no better than what we have now.

Identification experts call birth certificates "breeder documents," because they are completely insecure (most states will send you a copy of anyone’s birth certificate in the mail) and can be used to obtain ("breed") others. Driver’s licenses aren’t much better.

So let’s say that we get serious, and require a greater degree of proof: driver’s license, birth certificate, and the oath by two witnesses that the person is who he says he is. This will, at least, require that terrorists operate in threes. It also means a much higher burden. If filling out the applications and going to the Post Office, or wherever, to get a National ID takes each person, on average, two hours (a very optimistic estimate, as anyone who has been to the DMV can attest), we’re looking at 560 million person-hours, or the equivalent of over 900 lifetimes.

So in a best case scenario, we will have wasted the equivalent of 900 lives at the outset of our National ID program. Add in the effort required in processing, renewing, and following up on these ID cards and the toll in lost lifetimes could easily exceed the number who died in the Sept. 11 attacks.

The transition to a National ID would be painfully difficult for those on the other side of the window, too. If 280 million people need National ID cards, who will process them? In this quantity, it won’t be the folks who do security checks for the military and intelligence agencies.

They may not be perfect (can you say Aldrich Ames? Robert Hansson?), but they take weeks or months to clear people. With 280 million, about the most we can hope for is a careful review of the documents, a short face-to-face interview, then a check against a computer database.

Let’s say that this takes fifteen minutes (the two-hour time for the applicants still holds — they’ll spend the extra time filling out forms and waiting in line). Working seven hours a day (these are federal employees, remember), each employee could process 28 ID applications per day. That means that a million employees could process 280 million applications in just 280 workdays — let’s be generous and call it a year.

Of course, we don’t have a million full-time federal employees to do this (if the federal employees we have now can spare this much of their time, we should be laying them off in carload lots), and we can hardly afford to hire so many just for this job. (A million employees at $30,000/year is $30 billion/year, and that doesn’t count benefits, etc., that would likely push the cost above $50 billion, without allowing for the non-trivial expense of interviewing,and doing some kind of security check on, each candidate). So let’s do it with fewer: a mere 100,000 federal employees.

Of course, then it would take ten years to issue all the National ID cards. (And this doesn’t allow for the millions of new Americans who would be born or naturalized during those ten years). The total cost, of course, would remain the same, just spread over a longer time. But, frankly, if we can wait years for the antiterrorism protection that a National ID card would bring, then we don’t really need it anyway.

The other way we could reduce the the cost would be to have the federal employees spend less time on each application. Cut the time to three minutes each and you could do it in a year with a mere 200,000 federal employees, or ten years with a manageable 20,000. Of course, three minutes isn’t long enough to determine whether people really are who they say they are, so the result will be that the National ID will provide next to no security.

Add to this the certainty that some people involved in processing the documents will be corrupt or corruptible (or even terrorist sympathizers) and even a successful transition to a National ID system would leave fake documents readily available. How many will terrorists need? There were only 19 hijackers on September 11, and most of them entered with genuine, legal documents using their actual names. National ID wouldn’t have stopped them, but how hard would it be to get 19 fake ID cards out of 280 million? Considering the size of the investment, we’re not looking very secure.

A National ID scheme is a great deal for Oracle, and for the businesses that hope to use this information to send you more junk mail or assess your creditworthiness (and you can bet they will, whatever promises are made at the outset). It may also prove of some modest benefit in chasing down garden-variety criminals who aren’t smart enough to get fake documents. But for $50 billion and tens — or hundreds — of thousands of federal employees, we could do a lot more to promote security against terrorism. How many spies, law enforcement agents, special forces units and smart bombs will $50 billion buy?

You don’t beat terrorists with broad, diffuse, easy-to-penetrate defenses. You beat terrorists by killing them, by depriving them of bases, and by punishing nations that support them. For $50 billion, we should expect a lot more antiterrorism than we’ll get from a National ID card.

Glenn Harlan Reynolds is Professor of Law at the University of Tennessee, and writes for the InstaPundit.Com website.