On 25-jul-2006, at 22:50, Mark Smith wrote:
Along those lines, I'm curious what you (and other people who seem tobe against /48s for end sites) think of the "excessive" 46 bits of address space that ethernet uses, when the reality is that no morethan 12 bits of address space would probably have been plenty for theeven the biggest LAN segments
I could probably say a lot about the MAC addresses but this is the wrong SDO.
Fair enough. I was really only trying to prompt thinking about what simplicity and operational advantages have been gained from using "excess" address space in ethernet, when that amount of address space certainly wasn't necessary around 1980
Ethernet originally had 16- and 48-bit addresses. The 16-bit addresses apparently didn't catch on. The advantage of having a long address is that you can burn in a unique address in the factory and be done with it. With 16-bit addresses that wouldn't be possible, and with 32-bit addresses it would have been difficult given the u/l and g bits and the two-level hierarchy.
and, based on the address management tasks people are performing with IPv4 successfully enough today,
People thought coal stoves worked well too back in the day. But generating the lower bits in the IPv6 address from the MAC address is so much more convenient that having to administer IPv4 host addresses manually or through complex automation seems like stone age technology in comparison.
wouldn't be necessary now either. Operational and functional convenience was prioritised over necessity when the ethernet address size decision was made, and has paid many dividends.
I'm sure that if ethernet addresses had been managed by the RIRs we'd have EtherNAT today. :-) :-(