Imagine showing up for work and discovering you’ve got 457 voice-mail messages. Almost all of them are telemarketing junk — but somewhere in the mess, there’s probably a few important voice-mails from actual co-workers. How long would it take you to listen to them?

That may soon become a reality for many people, because of “voice over IP” phone numbers. Why? Well, as you probably know, VOIP routes phone calls over the Internet as packets of data, instead of using normal phone switches. That means VOIP phone companies can offer steep discounts on long distance. Indeed, that’s why many corporations are turning to VOIP — to cut costs! Even better, because the phone number is entirely, VOIP can do some cool tricks: Your number can be directed to any VOIP number in the world, so it can travel with you. And you can access your voice mail by checking a web page, because your voice mail is, of course, just data stored on a server, like email. VOIP will only get cooler and cooler as time goes on, because new phone tools can be rolled out at the speed of software.

But here’s the incredibly depressing downside: voice-mail spam. Since a VOIP phone number is just a locator to an online address, a spambot could automatically deluge VOIP voice-mailboxes with telemarketing calls. Indeed, a few VOIP users have already started receiving this stuff. There’s a terrific story on CNET about this, and the reporter discovered something alarming:

Consider software from Frederick, Mass.-based Qovia, which seeks out the IP addresses assigned to phones, then sends each a 30-second recording. Its pace — 1,000 synthetic calls every five seconds — is a quantum leap from the automated “Demon Box” dialers that telemarketers use now.

Qovia Chief Technology Officer Choon Shim said the company didn’t create its VoIP spam generator to send “30-second calls about Viagra to millions of phones.” Rather, it was to serve as a wake-up call of what could be a devastating problem for the growing Net phone industry, Shim said.

I have no idea if the VOIP companies are actually taking this seriously. But even assuming they are, I have no idea how they’re going to tackle it. Gad, look at how hard it is to stop old-fashioned email spam. We’ve spent years developing Bayesian filters, blacklists, and IP-activity sniffing tools, and we’re still getting flooded with the crap.

What’s worse, audio is infinitely harder to screen than text. Plain-text email is simple and easy to scan; but disambiguating the content of a complex spoken-word audio message is, like, intergalactically hard, as anyone who’s tried to use a voice-recognition system to write a letter would know. I am kind of fascinated to see how bad this problem gets.

(Thanks to Slashdot for this one!)