variable length addresses are a better idea than it appears at first glance. they do bring certain difficulties with them, especially when trying to do fiber-speed switching in hardware.
Poppycock. Hardware for switching variable length addresses first showed up about 15 YEARS ago. This is Not Hard at all.
it's hard to switch on arbitrary bit boundaries
Hello, we've been doing that since CIDR.
and it's harder to switch in hardware using prefixes of arbitrary length.
You mean addresses of variable length. One way: when you parse the packet, stuff the length (in bits) of the address in a register. Have your hardware autodecrement this register every time it matches a bit in the lookup process. Store any intermediate result in a fixed, known location. If the number of bits remaining in the register hits zero before you match a prefix, then promote any intermediate result to the final result.
overall it strikes me as a hell of a lot better than NAT.
Please visit the appropriate parties with your time machine. Immediately after the Big Ten meeting would be a fine start. Tony _______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf