There is a big change coming soon to Internet connections at your house and mine. Well, at least it is coming to my house. I'm talking about a significant increase in bandwidth for not such a significant increase in cost. My Internet connection is about to be upgraded to a Digital Subscriber Line (DSL), and I can't wait!

DSL is a digital telephone connection that is somewhat like ISDN but 3-10 times faster. It comes in a variety of speeds depending on how far the user is from the telephone central office. Any connection within 18,000 feet of the CO is supposed to be good for up to 1.5 megabits-per-second, which is about the speed of the T-1 lines bought by most Internet Service Providers. Unlike ISDN, DSL lines generally don't have per-minute charges and are on all the time -- 24 hours per day seven days per week, which opens up a world of possibilities I'll discuss in a bit.

DSL comes in many shapes and sizes. Some of them allow faster downloading than uploading because most Internet users download more dirty pictures than they upload. Asynchronous Digital Subscriber Line (ADSL) is one version that offers 1.5 megabits-per-second downloads and 64 kilobits-per-second uploads. But the connection I am getting on February 23rd is symmetrical -- 1.1 megabits-per-second in both directions, replacing the existing 128 kilobit-per-second ISDN connection that I will keep as a backup.

DSL was an invention of AT&T Bell Labs and Bellcore. Bell Labs, of course, was where the transistor was invented, but Bellcore is a less well-known research lab that's funded by the Regional Bell Operating Companies. After the breakup of AT&T in 1983, the RBOCs felt a need to continue some basic research, and that happens at Bellcore. The DSL work was done in the late 1980s, but it's only reaching the market now because consumers are demanding faster connections than ISDN, and because telephone deregulation has made it almost impossible for the phone companies to continue keeping DSL under wraps.

Today, in most telephone markets, new companies can declare themselves to be in the local phone business and start wholesaling local lines from the traditional local phone companies. In my case, here in Silicon Valley, the old phone company is Pacific Bell and the new phone companies are a bunch of outfits we've never heard of. But for about $10 per pair of wires, the law says they can buy local lines from Pacific Bell. And having bought those lines, they can offer over them any kind of service they want as long as that service will run on a plain old pair of wires. If you were starting a local phone company, would you try to resell those local lines for voice use at anywhere from $13-20 per month or for DSL at anywhere from 3-10 times that price? Right, you'd offer DSL, and users like me would be thrilled to pay that amount.

Of course, the traditional phone companies don't like this at all, and one way they try to minimize the problem is to be very stupid about actually handing over lines to these upstart local phone companies. Another way is by claiming that most local lines aren't good enough for DSL. Very often this is not true, but the traditional phone companies don't want us to know that. The truth is that most local lines are perfectly good for DSL, but you have to test them first. My lines went in the other day and tested fine at 1.1 megabits-per-second. The installer just grabbed a spare pair of wires in the box on the side of my building and I was in business. Well, almost in business, since I have to wait until the 23rd before my ISP is up and running at that speed.

What can you do at speeds above a megabit-per-second? Well, what I am going to do is install a server to bring a few more bells and whistles to this Web page. Look for announcements on that toward the end of the month. But running your own server is the least of it. A 1.15 megabit-per-second connection is enough to run fullscreen video CDs (NOT DVD, but the lower-quality video CD standard that is popular in Asia). An asymmetrical connection that allows a 1.5 megabit-per-second download is fast enough for fullscreen MPEG-1 video, which is equivalent to VHS. So you could rent movies over the Net if only the Internet backbones could support that level of traffic, which they won't be able to for several more years.

So the truth is that if millions of people think they are going to be downloading like crazy and watching "Terminator 2" at 1.5 megabits-per-second, they are wrong, wrong, wrong. The system won't support it. You'll have that kind of connection to your ISP, but not upstream of the ISP and across the Internet to the really good stuff. Of course this makes no difference to me, because my goal is to serve PBS Online customers who are generally logging-in at 28 kilobits-per-second, but if I was upgrading just to watch dirty movies, I'd feel cheated.

So what are we going to do with that extra bandwidth? I have a plan, a revolutionary plan, to use DSL lines to take over the Internet.

There was a time not that long ago when the Internet was closed to all but a few thousand users at major research universities, defense contractors, and of course the military, who paid for it all. This was in the days of ARPAnet and later, the NSFnet. Things got more liberal after 1988. Until that time, most companies, most schools, and nearly all individuals were prohibited from using the Net. So they built their own.

These misfits invented the Unix-to-Unix Copy Protocol (UUCP) to create an unrestricted network among Internet untouchables -- the USENET. USENET was used mainly for electronic mail and newsgroups. In fact, the newsgroups we know today were born of UUCP and not the "real" Internet.

USENET had no backbone in the sense that the Internet has backbones. Instead, every participating institution was responsible for buying and maintaining a single link to another institution. This could be a leased T-1 line between one school or company and another, or just a dial-up telephone line. The basic tenet of the UUCP was that data was stored and forwarded by these nodes until it reached its final destination. Each member institution got access to the transfer capacity of every other institution and the only charge was quid pro quo: I'll carry your traffic if you'll carry mine. And it was fair, too. Institutions that paid for big lines had the capacity to send and receive more data than institutions with small lines. Everybody got their money's worth.

But UUCP was far from perfect. For one thing, it was slow, especially around the edges, since some of the smaller connections weren't up and running full-time. This was okay for e-mail, but not okay for Web browsing, except there was no Web browsing back in the mid-80s. UUCP also didn't have automatic routing like the Internet. You not only had to specify a destination for your e-mail message, you had to specify the route to be travelled to reach that destination. This was done by using long addresses that included routing information separated by exclamation points, which were called "bangs," when you read your address out loud to someone over the phone.

UUCP went out of favor after 1988 when the National Science Foundation finally allowed commercial users on the Internet. With faster links and automatic routing, the Internet was much better than UUCP. But now I think UUCP's time may have come again.

So I propose a new UUCP based on DSL connections. In this scheme, every participating host has to have a pair of DSL connections. Initially these would have to be to different Internet Service Providers, but in time they could just be to different telephone central offices. That's because this new UUCP would eventually kill ISPs as we know them. Why kill the ISPs? Well for one thing, I don't think they do a very good job and they cost too much. But wouldn't this put all the power back with the phone companies? Not if they were just suppliers of DSL connections at $10 per month. The phone companies would be so poor so fast that they'd soon be able only to serve the public -- there's a novel idea. And instead of having an Internet with 29 commercial backbones averaging about 60 megabits-per-second each (about 2 gigabits-per-second total) we'd have 20,000 virtual backbones of around one megabit-per-second each. That's 10 times the bandwidth for less money. And what would we do with all that extra bandwidth? We'd carry voice and video and all the other services that we'd be paying extra for a decade from now except that this way we wouldn't be paying extra at all. As an anarchic network, it would be impossible to bring down, too.

Of course, there will be horrific routing and network management headaches to be dealt with in building a network like this, but I figure those Linux programmers need a new challenge.

My Mom calls this "Godless Communism," and that her AT&T stock is low enough as it is. Maybe she's right, but still I think it would be fun to try. I'll pop for an extra DSL line or two, how about you?