kwe@bu-cs.BU.EDU (kwe@bu-it.bu.edu (Kent W. England)) (05/13/88)
In article <8805100150.AA07641@braden.isi.edu> braden@VENERA.ISI.EDU writes: >I disagree. The Internet standard approach for a host: pick any >gateway and let it redirect you... is simple and effective. You REALLY >DON'T want hosts to know about routing, even subnet routing!! I agree in principle, but in practice we are talking about hosts on Ethernet LANs with flat address space internetworked on IP wide area networks with hierarchical address space. I don't see any alternative to ARP and gateways. The host has to know when to look directly on the flat address spaced LAN and when to hand off to the gateway for forwarding. This brings me to a question I have been dying to ask for a long time, but just didn't quite know what context to ask it in... It seems so obvious to me that a hierarchically structured address space in Ethernet (read 802.x) 48 bit addresses would be a great improvement over the current kludged 48 bit flat vendor-assigned address space coupled with the 32 bit IP address space that I wonder why the issue never comes up. I know all the obvious problems with structured Ethernet addresses, but everytime I look at the issue it seems to me that structured Ethernet/IP (read 802.x) addresses would be a great improvement over flat address space. I think the 802 spec allows structured addressing, yes? Why is there no hint of movement in this direction? Waiting for ISO? Or have I failed to grasp some essential difficulty with this? Comments, please. Kent England, Boston University
romkey@kaos.UUCP (John Romkey) (05/15/88)
In article <22596@bu-cs.BU.EDU> kwe@buit13.bu.edu (Kent England) writes: > I agree in principle, but in practice we are talking about >hosts on Ethernet LANs with flat address space internetworked on IP >wide area networks with hierarchical address space. I don't see any >alternative to ARP and gateways. The host has to know when to look >directly on the flat address spaced LAN and when to hand off to the >gateway for forwarding. Easy. If your host has only one network interface, then just do this: if((my_ip_address & my_subnet_mask) == (dest_ip_address & my_subnet_mask)) first_hop = dest_ip_address; else { first_hop = default_routers[next_default_router++]; next_default_router %= number_of_default_routers; } Then send the packet to the IP address on your physical network given by 'first_hop'. > It seems so obvious to me that a hierarchically structured >address space in Ethernet (read 802.x) 48 bit addresses would be a >great improvement over the current kludged 48 bit flat vendor-assigned >address space coupled with the 32 bit IP address space that I wonder >why the issue never comes up. I know all the obvious problems with >structured Ethernet addresses, but everytime I look at the issue it >seems to me that structured Ethernet/IP (read 802.x) addresses would >be a great improvement over flat address space. I think it would be a bad idea to give up the 48 bits of addressing. 48 bits is more than we'll ever need, right? Fine, then let's keep it that big. *Maybe* we'll really never need it bigger (yes, I can't remember if 2^48 is greater than the estimated number of particles in the Universe...if it is, that's even better). So let's not take any of the current 48 bits to form the hierarchical address. The only reason that I think you'd really want to have hierarchical ethernet addresses is if you want to use them as your IP level 'logical' addresses, but if you do that then you open up lots of problems: the addresses must be variable length if you want to support more media than just those that use ethernet-style addresses, which is really nasty from an implementation point of view (like ISO, which in general is really nasty from an implementation point of view); and you end up with volatile addresses wired into IP-level addresses which have a (presumably) long lifetime - when your ethernet interface goes to bit heaven, you don't want your computer's IP address to change. -- - john romkey UUCP: romkey@kaos.uucp ARPA: romkey@xx.lcs.mit.edu ...harvard!spdcc!kaos!romkey Telephone: (617) 776-3121