Two years ago, 3Com dramatically changed its networking focus, and many commentators thought they had seen the last of it in the enterprise space. At a stroke, it discontinued the big CoreBuilder switches that had been its main product for the heart of the corporate network, and handed the customers over to Extreme Networks. In the following months, the company floated the Palm PDA business it had acquired with US Robotics, and also hived off its modem business, effectively spitting out the rest of US Robotics. It all looked like the typical dot-com deconstructions of the time. Broad-spectrum but unexciting network companies like 3Com were routinely torn into pieces. The bits that could be made to seem "cool", because they dealt with service providers and other then-exciting markets, would hope for astronomical stock valuations, while the rest -- divisions dealing with dull sectors like the enterprise -- could sink or swim. Like so many other dot-com restructures, it wasn't a great success. 3Com's DSL and telecoms ideas did not come to the fruition that was expected, though the company puts a bright gloss on this: "We were innovating ahead of the curve," said Karen Oddey, 3Com's vice president of marketing and business strategy. "Then the economy thing happened." The most surprising thing about all this is that, two years later, the company is still in the enterprise sector, apparently doing very well, thank you. How come? Oddey's answer has the air of an oft-repeated explanation: "We never left. We actually exited the market for large complex chassis switches, often based around ATM, and shifted to Gigabit Ethernet." This is more than a little disingenuous. Extreme was the doyen of Gigabit Ethernet companies and would not have taken those customers if they could not be moved easily from ATM. So it seems strange that 3Com couldn't be bothered to do that itself. But all that is in the past, and the future is much more interesting to a strategist like Oddey. She is much more keen to talk about 3Com's next big push in enterprise territory, which will use the cheap modular Ethernet switches it continued with, to replace the big chassis switches it ceded to the opposition two years ago. The move, called XRN, was described in detail at Networld+Interop last week. XRN adds redundant components to stackable switches, and promises links that will let a bunch of them operate as one geograhically distributed switch. Like so much else from the network industry, industry, it's a roadmap of something not actually available now. 3Com followers will remember similar promises: last year, the company spoke of a personal Bluetooth hub that would get the cables off the business desktop, and before that, it promised NICs made intelligent by embedded Microsoft Windows 2000. I wondered why some of 3Com's ideas happen and some don't