Well, all that has changed. The sleek new switch-routers use custom ASICs to increase performance while holding down price. They also typically limit the sort of hardware and protocol options from which you can choose. Cisco and Bay offered the ability to route everything from Token-Ring to smoke signals. The new guys will let you do anything you like, as long as it's some form of Ethernet. The old routers would still handle obscure stuff like Apollo Domain along with any tiny twist to TCP/IP and IPX. The new boxes do IP and IPX, and that's it.

The net result is a very fast and very affordable router that probably will do everything you'd ever want it to do within a corporate network. Imagine getting Gigabit Ethernet ports for the same price you used to pay for 10-Mbps Ethernet ports. That should help corporate IT's bottom line. You'll move tens of millions of packets each second for the same price that used to move tens of thousands of packets. That's hard to beat.

Of course, what if you have other protocols? There's still a lot of AppleTalk and those silly Microsoft protocols floating around. What do you do then? Luckily, these boxes can be configured to

either bridge or ignore protocols that aren't understood at the network layer. This lets conventional routers back up the new hot routers and handle protocols that the new box can't. Hopefully, this will extend the life of your older router, and maybe by the time it really does wear out, you'll have transitioned all your apps and users to IP.

Caveat Emptor! It all sounds great, and there is a slew of proposed and ratified standards that makes it sound like a less-proprietary solution than many of the other Fast IP solutions some vendors are proposing. The question is whether sinking various standards in silicon will make these boxes obsolete before their time.

If you're thinking about buying some of these new wonder-routers, be sure to ask some tough questions. Ask for assurances th at protocol updates will be supported. Bandwidth reservation and prioritization schemes are especially important. This stuff often has to happen in hardware, so if these issues are important to you, get support guarantees--in writing. Joel's Gigabit Ethernet review gives a good list of protocols about which to ask. If IPv6 is important to you, make sure you get the dirt on that, too.

ATM is DOA. Forget MPOA. These new products and the slew that will be introduced over the next few months will definitely change the back end of network infrastructure, just as the super-low-cost Ethernet switches have changed the way we do networking at the edge of the network. By this time next year, we'll be talking about VLANs and Gbps uplinks. NIC design and I/O subsystem throughput will again become very important. Servers will strain to fill their high-speed network pipes.

It's a great time to be designing network infrastructure. By this time next year, Ethernet will once again be king. We'll wonder wha t the fuss over ATM was all about. Watch for Bay, Cisco, 3Com and all the start-ups to vie for your imagination and dollars, and watch ATM slowly fade back into the telco realm where it belongs.

Art Wittmann can be reached at awittmann@nwc.com.