But I felt that Fast Ethernet was plenty fast for our backbone--in fact, it was just about as speedy as the ATM OC-3 solution, once you factored in the cell tax. And I believed Fast Ethernet would be easier to implement and would require less training to support. I was sure there was nothing to stop Ethernet from getting even faster down the road. Still, Fore's mantra, echoed by one vendor after another, nagged at me, causing me to doubt my position. It didn't help that our colleages at nearby Cornell University were going all out with their ATM-to-the-desktop installations and leading research efforts to make the transition easier (remember "cells in frames"?). We didn't want to be left in the dust while the rest of the world was on the cutting edge.
In the end, though, my peers and I refused to succumb to the pressure of public opinion, and in holding steadfast we avoided a drastic (and costly) detour that forced some of our colleagues to rip out their ATM installations and replace them with Ethernet.
Steer Clear of Vendor Predictions
Vendors are still trying to lead us in the wrong direction. In the '90s, Cabletron's motto could have been: "All roads lead to Layer 2 switching." Cabletron didn't have a routing solution back then, either. Now many vendors have gone to the other extreme, putting Layer 3 in wiring-closet switches. Do "all roads lead to routing in the wiring closet"?
You may have some of your own "All roads lead to insert latest vendor-touted technology here" examples. How about "All roads lead to IPv6"? Actually, all roads might lead to IPv6 eventually, but it's going to be a long, bumpy ride, and most of us have nothing to gain by climbing on board early. It could take up to 20 years for IPv4 address space to run out (see Geoff Huston's column, "IPv4--How Long Have We Got?"), thanks to CIDR and NAT. You may find some vendors along the way who will try to convince you to think otherwise--and, coincidentally, you may also find that those vendors have IPv6 solutions. Do not give in! Let the academics and the Department of Defense pave the way, and wait another couple of RFP cycles before even considering IPv6.