How high can you clock a 68K FPGA implementation? That to me seems far more interesting than an x86 accelerator or remote execution / remote procedure call enviornment. The ColdFire is cool, but can you make a faster, more compatible processor in an FPGA? Sorry, I haven't been following the FPGA threads.
The idea of using an x86--or any other host system--for remote execution certainly isn't anything new. Rather than move GUI data back and forth, you're better off having a local GUI and just transporting data and procedure calls. Or something like that. High-speed interconnects would be prefferable--maybe GbE on an accelerator card with an appropriate buffer ("instructions" and data) and custom (or open) protocol? It would then at least be interoperable with most of what's out there today.
