Capitalists, Socialists, and ATM Switching

Peter Newman
Data Communications Magazine, Dec. 1994

There are two fundamental philosophical approaches to switching: capitalist switching and socialist switching.

Capitalist switching goes something like this: You give me some money and I give you some bandwidth. It's all yours. You don't have to share it with anyone. If you don't use it, then it is wasted, but hey—that's capitalism.

The capitalist model persisted for well over half a century, and served us well for the traffic we were switching at the time. But during the 60's there arose a new school of thinking—socialist switching. It was made popular by the People's Republic of Berkeley (home of the free Unix movement) and caught on in part because it was hip, but mostly because it was free. In socialist switching we have a huge pool of bandwidth—and we share it! You want some bandwidth, man, you take it, and when you're done, kindly return it to the pool. Revolutionary! But now we encounter some fundamental problems endemic to the socialist system—there is never enough to go around. So we all stand in line at the bandwidth bottleneck awaiting our turn for transmission. But who gets to go first? To solve that we need red tape (you were expecting some other color?) in order to share the resources. In the trade we refer to it as protocol.

Thus we have two approaches to switching. And behind these two philosophies grew two huge industries: the telecommunications industry, with a staunchly capitalist model of communication, and the data communications industry, which is generally much more hip.

Now around the time that people had bought just about as many phone lines as they needed, the telecom industry got to work on what to do next. The answer was integrated services. Unfortunately, these bore such a resemblance to a pair of phone lines that they failed to catch the customer's imagination, or pockets—even though they were very nice (digital) phone lines. So the capitalist switchers bought themselves a marketing department to tell the customers exactly what they needed. And they started some work on how to integrate the services themselves, not simply the access to the services. Thus began some really cool research into new switching mechanisms capable of much higher capacity and far greater flexibility—the first small step down the slippery slope to socialism.

The problem with the capitalist approach (circuit switching) was that you could only get bandwidth in discrete lumps—56 kb/s, 64 kb/s, 1.544 Mb/s, 2.048 Mb/s, and so on. It didn't take long for the marketing department to figure out that there were other bit rates, and some of them were really high. And maybe you could do really cool things with those higher bit rates, like multimedia, which was just around the corner. But it was getting increasingly tedious to attempt all this with circuit switching.

Meanwhile, back at school, the socialist (packet-switching) folks also were hard at work. They had hit upon a truly astounding insight: all you really needed to do to make things go blindingly fast was to throw away all of this protocol business. They already had all the flexibility they needed so they were concentrating on high bandwidth and the ability to integrate delay-sensitive traffic (multimedia applications being just around the corner, simply waiting for the really high bit rates).

So without looking where they were going, the capitalists pursued the flexibility offered by the socialist approach, while the socialists were after the real-time capability of the capitalists. And they collided head-on in the middle. The capitalists called it ATM (asynchronous transfer mode) the socialists, fast packet switching (the same thing, provided you use short, fixed-length packets). Thus ATM became the first switching technology capable of supporting both capitalist and socialist switching within a single integrated switching mechanism. And nobody realized what a colossal blunder had occurred.

You see, until this time the telecomm industry and the datacomm industry were completely separate. They drank in separate bars (except for the guys in the blue suits, who mostly kept to themselves), and they didn't need to agree on anything more complicated than a few physical layer interfaces.

But now it's too late. Now we have the ATM Forum—the United Nations of the technical world. Now the capitalists and the socialists are not merely talking to each other but trying to hammer out a brave new networking world (that is, when they're not simply hammering each other). Thus far we've managed to integrate the technology. But will we ever integrate both sides of the industry? As we've seen, even longstanding walls can come down. But will it happen in our lifetime, or will it require a new generation—of engineers, not just technology?