The Layer 1 barons can come up with legitimate claims that interesting processor tasks have turned from packet forwarding to channel aggregation, a genuine trend that plays to the strengths of PMC-Sierra Inc. or Applied Microcircuits Corp. (AMCC). But far too many babies have been thrown out with a tanker car's worth of dirty bathwater this summer.

Admittedly, the cutbacks began long ago. PMC's Bob Bailey took the medicine earlier than competitors, eliminating large-scale programs in Internet forwarding architectures in the early summer of 2001. As a result, PMC was able to address expanded legacy Sonet markets more quickly than competitors-but at what price? The Canadian company got rid of some of its best executives and some of its most interesting architectures in the mad dash to cut costs.

This summer, Vitesse Semiconductor and AMCC joined the stampede to drop the packet-pusher chips. The Vitesse scenario was particularly frustrating, since it involved decimating the Colorado operations of SiTera and the North Carolina business of Orologic-despite the fact that both acquisitions were good ones for Vitesse, at least architecturally. AMCC has kept more of its MMC Networks and Cimaron acquired businesses, but the size of its layoffs may limit how effectively future processors can get to market.

Passing the packet baton to those most familiar with microprocessors doesn't mean they'll own the NPU market as business picks up. In fact, rumors abound that Intel is still interested in unloading good chunks of its communication business. But what is clear is that the physical-layer players had a chance at Layer 3 logic dominance, and punted. What was that Kenny Rogers said about holding vs. folding?