Tuesday, May 01, 2007

If I Could Redesign the Internet

If I could redesign the Internet, I'd fix it so I could find out who sent anything I receive: personal emails, spam, bomb threats, you name it.

If I could redesign the Internet, anybody who wanted to send thousands of emails at once, legally or otherwise, would have to pay up front first.

If I could redesign the Internet, my browser couldn't be taken over by some little ad for low-interest mortgages that suddenly balloons out and hides the thing I'm trying to read.

All right, so the last one is more along the lines of a pet peeve. But the first two are reasonable. If we could easily and reliably find out exactly who is sending spam and malware, it stands to reason that not nearly as many people would do so. And if bulk email had a cost structure similar to direct snail mail, spam wouldn't go away, but we'd get a lot less of it. So why don't we just fix these problems right away? The reason can be illustrated by a little story from my days as a radio engineer with a large mobile-radio firm, back in the 1970s.

At the time, I was on a team of fresh young engineers charged with designing a new mobile radio for police cars and fire trucks. One of the first things we did was to take a look at the connector between the radio and the antenna. The connector is like a bridge that carries the "traffic" of the radio waves. If the bridge is bumpy or full of holes, you're not going to get much traffic across the bridge. Similarly, if the connector is of poor quality, you're going to have problems sending the radio waves back and forth to the antenna. The connector on the old radio design we were replacing was called a "PL-259," a type that dated back all the way to World War II, and we decided we were going to replace it with a newer design that presented a smoother path to the waves. Then we had our first progress meeting.

At the meeting, an old-time manager listened patiently as we presented our ideas for the new design, including our plans for the new connector. "Are you finished?" he asked. When we said yes, he replied, "You kids obviously haven't heard about the First Commandment of mobile radio design."

No, we guessed we hadn't. What was it?

"Thou shalt only use a PL-259. Neither shalt thou even think of using any other connector." He pointed out that thousands of police cars and fire trucks all over the world had antennas that connected with a PL-259, and there was no way he was going to let us change it. It was what engineers call a "legacy problem": there's too much hardware (or software) out there that a change would obsolete. Thus perished the notion of updating the connector, at least for that new design. Eventually, long after I left the company, I learned that they did manage to replace the PL-259, but probably only after a long internal battle and a lot of hand-holding for customers who had to replace antennas or use adapters.

This minor episode illustrates the major problem with changing certain features of the Internet. Take the problem of anonymity. Way down at the level of the basic protocols or rules followed by all the machinery that runs the Internet, there is simply no way to ensure that you can figure out who sent what. The reason for this is partly historical. In the Internet's early days, it was a research toy shared by a few large, sophisticated, and trustworthy computing centers. For several years, it probably never entered the mind of anyone involved that one of the users would deliberately try to misuse the system to conceal their identity. By the time the Internet was large enough to attract such people, it was too late to start over with a new set of protocols that contained built-in security. There are also a lot of problems and delays caused by the fact that people using the Internet move around a lot now, with laptops, PDAs, Internet-capable cell phones, and whatnot. The system was originally designed to deal with fixed mainframe computers that were as likely to move around as the Washington Monument, and the patches and fixes that have been added to deal with mobile users are inefficient and complicated.

More patches and fixes aren't the answer. For these basic legacy problems to be solved, it looks like we will have to wait for a new Internet altogether. The National Science Foundation is paying for research into how we'd like such a new system to look with its Future Internet Network Design program (FIND). But estimates for how much it would cost to scrap the existing system and install a new one range into the many billions of dollars.

Who's going to pay for it? Well, one way or another we already support the present system, through bills to our Internet service providers, tax dollars, and other ways. It will be interesting to see how far we can stretch the old protocols, but some day they'll start looking the way that PL-259 connector looked to us young engineers. Right now it's not just a crusty old manager stopping us; it's the expense of changing over. But as the Internet becomes a vital part of life-critical services such as medical telecommunications, we may have to start something like a two-tier system, rather like the HOV lanes on freeways: an expensive but super-reliable and super-secure network, and then the regular old system for everybody else, with maybe nodes here and there connecting the two.

I'm no computer scientist, so I'll let the experts figure out how to make the transition. But spamless email and freedom from malware seem like pretty attractive goals, even if it does cost a bundle. And if somebody does eventually figure out a way around the new safeguards, we might have a few years to enjoy the Internet as it was intended to be.

Sources: A series of articles by Anick Jesdanun on redesigning the Internet was carried by the Associated Press and reprinted in several newspapers, and carried online in part by the Hartford Courant at http://www.courant.com/business/hc-rebuildinginternet.artapr15,0,5625095.story?coll=hc-headlines-business on Apr. 15, 2007, and also in the Austin American-Statesman print edition of Apr. 23, 2007, pp. D1 and D4.

No comments:

Post a Comment