Ghosts in the Marketing Machine

Peter Newman
With illustrations by Hans Bjordahl

Data Communications Magazine, 21 May 1996.

Welcome to the machine—the ATM Marketing Machine. The great ATM Marketing Machine (AMM) is a virtual machine. You feed it reality and out comes virtual reality. Virtual reality is cool, way cool. It's almost indistinguishable from true reality. Virtual reality is what reality will look like after the next roll of the die, after the next rev of the spec, after we deliver the next generation.

No fundamental laws are broken by this transformation of reality. The process operates within the confines of the Second Law of Market Dynamics: "All markets tend toward disorder, especially when acted upon by a standards body," as well as the Principle of Market Inertia: "All markets continue in the same direction unless acted on by an exterior cash flow." (In some cases, a huge exterior cash flow.)

Let me demonstrate the great ATM Marketing Machine in action. Feed in local-area networks and out come virtual LANs. Not that anyone has yet found a use for virtual LANs. They seem to be an excuse for not having fast, low-cost routing with auto-configuration. But virtual is cool, way cool.

Try another one. Feed in Ethernet and out comes LAN emulation—almost as fast as 100-Mbit/s Ethernet, as long as you don't try giving it any multicast traffic (although I wouldn't try squeezing the driver into boot ROM). And guaranteed to offer hours of fun to the IS department as it whittles away those dull weekend hours attempting to configure it.

Now this is all well and good and perfectly normal. The real problem came when someone perverted the process. I think an engineer must be to blame. One of those rare misfits who traversed the chasm to marketing but became embittered when he didn't get invited to those exotic places where the cool, way cool, standards people go.

Instead, he was left behind to feed reality into the great ATM Marketing Machine. Being very bright, but somewhat lazy, he figured that if he took the virtual reality output from the great ATM Marketing Machine, delayed it just a tad, and then fed it back into the input he wouldn't need to stoke the Machine with reality any more—saving considerable effort.

In so doing he achieved the legendary Perpetual Marketing Machine (PMM). As long as the cash flow is maintained the output is sustained. It gets a little thin on reality after a while, but the strange thing is when everyone returned from their meeting at one of those wild and exotic places, nobody noticed the difference.

Now I realize all this sounds a little farfetched, but how else can you explain the emergence of the Cells In Frames (CIF) proposal? Let's check that I have this straight. I take a packet, the sort I could wrap in an Ethernet frame and ship onto the wire using less than 2,000 lines of driver code. Instead, I feed it into 10,000 lines of LAN emulation client that my sys admin has configured (correctly, if I'm lucky). Then I ask for a VCI—using 40,000 lines of signaling code. And when I get one I shred the packet—remember the packet?—into lots of little cells and wrap them in an Ethernet frame.

Finally, I transmit the resulting packets-in-cells-in-frame over my Ethernet wire (the sort that could be carrying perfectly serviceable Ethernet packets) and send them to an ATM switch that is offering LAN emulation service. That's right—an ATM switch that is in fact pretending to be an Ethernet (and not a very good one at that). And I'm doing all this because integrated voice/data is the killer application?

It is indeed a killer application. The integrated voice/data path is littered with small corporate gravestones and not an insignificant number of less-than-amazingly-popular standards: FDDI II, DQDB, and IEEE 802.9, for example. If you really want integrated voice/data why not use iso-Ethernet? At least it doesn't rely on virtual reality.

Cells In Frames reads like an epitaph to me. I could be wrong. It would not be the first time. And maybe it is foolish of me to burn my bridges, but then we all know bridging is a limited solution. If you really want to fly, try IP switching.