In the airy realm of theoretical math, there's an infinite supply of numbers. But come down to earth and look at telephone numbers and you'll find that they're running short.

It's partly because of the millions of information-age playthings -- cell phones, fax machines, computer modems -- that have their own numbers. But you can make a case that it's also because of the telephone industry's practice of wasting numbers. Strange as it sounds, there are millions of telephone numbers that could be used, but aren't.

Last week the Federal Communications Commission voted to begin formal consideration of a "use it or lose it" policy for companies that get rights to numbers, then sit on them.

The goal is to postpone until well into the future the scrapping of our 10-digit system, with three digits for the area code and seven for the local number. Making us dial an extra digit or two would create a vast supply of new numbers, but require a very costly overhaul of the national phone system.

The shortage is more than a concern for network engineers. The faster that numbers are used up, the faster phone companies take the unpopular stop-gap step of creating new area codes. Each area code creates more than 7 million new assignable numbers, but the new codes also means big expense and bother as people change letterheads, reprogram speed-dialers and write to customers to say the old number is wrong. It also can mean people have to dial 10 digits for all local calls, as is the case in Maryland.

Area code shock has become one of the nagging realities of life in the late 1990s. There were 119 codes in the United States in 1991. Today there are 215 and about 70 are close to being maxed out. What's more, there are only so many area codes that can be added -- about 411 as of last count.

Phone numbers and area codes come in finite supply because there are only so many combinations that can be achieved with a given number of digits. And many of those combinations can't be used -- a local number, for example, can't begin with 911 because the phone system is programmed to ring the emergency line the moment those three digits are entered.

As the FCC and many in the industry see it, part of the problem is the no-questions-asked approach of the North American Numbering Plan Administration, a private industry group that oversees a number assignment system that was drawn up half a century ago. Lockheed Martin Corp. of Bethesda runs the program.

It's basically an honor system: If a telephone company comes in and says it needs new numbers, it generally gets them -- and in blocks of 10,000, for free, with no one inquiring as to whether they're really needed.

If you're a hot new cell phone company, you might run through that in a month. But if you're a company wanting to compete in the local "wireline" market -- those calls across town, a business now largely the preserve of Bell Atlantic Corp. in the Washington area -- you may end up with no use for a lot of them.

Partly that's because for all the promise we hear of competition in local calling, it has yet to bloom. The new companies remain bit players against the Bells, and many of their numbers sit waiting for customers. Not that this warehousing is entirely their doing: The Bell companies also have stocks of unused numbers.

Whoever's got more, one recent study determined that roughly two thirds of all numbers assigned to phone companies aren't being used.

Another reason for the waste has to do with the design of the phone system. Prefix numbers are associated with geographical areas and phone companies generally want to be able to sign up customers everywhere. So they need allocation all over an area code and they get those 10,000-number blocks. Never mind that in one little corner of a county they've only got 50 customers for now; the remaining 9,950 numbers will go to waste.

What the FCC did last week was propose that numbers be handed out in blocks of 1,000 and that it consider whether phone companies should pay for them. The agency also proposed watchdog measures to assure that the companies really do need the numbers they hold, and if not, that the numbers flow back into the pool for use elsewhere.

By some estimates, steps like these could extend the life of the 10-digit system by more than 75 years, though in this industry any prediction is a guess and nothing more.

The next big move would be to add an extra digit or two to every number. That's no easy task, however. It would be a huge software job. The telephone system is a series of very complex interconnected computers, and all would need extensive reprogramming.

Industry estimates suggest that such a changeover would cost $50 billion to $150 billion and take a decade to accomplish. That's not as expensive as fixing the millennium computer bug, but it starts to feel as if it is in the same league.

Showing a little sense in conserving what we've got sounds attractive now, doesn't it?

John Burgess's e-mail address is burgessj@washpost.com