If someone were to take *half* the software innovations which have been made over the past 15 years (a decent fib, interrupt coalescing, compiled packet matching rulesets, etc) and applied them as if they knew something about networking and coding, they could very easily produce a box using off the shelf PC hardware which woops up on a 7206vxr for somewhere less than $2000.

Do you have any evidence that these improvements are not being done? The people building supercomputer arrays using Linux have a need for consistently high pps and bps that is greater than anything we see today on the Internet. They've been working on these types of improvements in device drivers and the OS (Linux, *BSD) for years now. You might not find this stuff in a standard enterprise distro like RedHat or SUSE but it is trvial to source this stuff and integrate it into your own build of the OS. A lot of this discussion has been people guessing about performance issues but few people have taken the time to put together a few boxes with Linux or *BSD and either Zebra or Quagga to trial them. We all go through detailed evaluations when buying C or J boxes, so it's not a waste of time to trial some Z or Q boxes as well to see what they can do. In the end, the resulting performance is affected by so many factors that it can't be predicted without testing. For instance, any weaknesses in the software might be completely nullified by the greater CPU power of a PC platform. And lets not forget that there are other platforms like ARM and PPC. Here's an ARM development system with PCI http://www.simtec.co.uk/products/EB110ATX/intro.html and here's a PPC one http://www.artesyncp.com/products/PM-PPC-440.html And if anyone thinks that ASICs give C and J an speed advantage that others can't touch, then guess again. Nowadays those ASICs are probably programmable ASICs which is a fancy way of saying that they are mostly made up of FPGA cells. It is not that difficult or expensive for people to design and build their own ASIC using cheap FPGA technology from companies like Xilinx. This is basic sophomore level electronics and is simple and cheap enough that people even hack their own MP3 players using FPGAs http://www.pjrc.com/tech/mp3/fpga/ PC-based routers may not be magic bullets but I think we should take them a lot more seriously especially if you want to innovate and offer something that differentiates you from other network operators. In a world where everybody runs C and J networks, there is only one flavor available, vanilla. --Michael Dillon