No one seemed to heed this advice, as many startups were considering new ways to overbuild the telecom infrastructure in support of the Internet and World Wide Web during the late 1990s. Vinod Khosla, in particular, ignored this advice during 1997 as he nurtured his Fiberlane startup to produce an IP over SONET multiplexer to address the bandwidth tsunami he foresaw. Data traffic had to be embraced for carriage across the existing infrastructure, and it could best be achieved by retooling the optical infrastructure to carry both voice and data; and data carriage could be embraced in any format presented, be it IP packets, ATM cells, or Frame Relay traffic.
Vinod’s bold vision ultimately cost Cisco more than $7 billion dollars just three years later when the company felt it had to buy an optical infrastructure startup, Cerent, the progeny of Fiberlane, to address this market segment.
Maybe Ed should have looked closer at some of those infrastructure startup companies in the first place.
 March 1997 The Red Herring article, Routing the Competition (pp.82–88).