Hi Frank, a couple different issues were raised in your post, so I'll comment on one now, the others later.
>>re: "...the almost universal use of network
address translation (NAT), DHCP, etc.. where end users never wind up with the same
address twice. Very big problem ..."
It is a common misconception that users never end up with the same IP address with DHCP, but in fact you SELDOM get a different IP address. This is "by design" and is the specified primary behavior per RFC2131 (the dhcp "spec").
In general, the only time you would get a *new* IP address is if your IP lease reached the renewal phase of DHCP (which is still a long ways away from actually being "expired") and the DHCP server which originally provisioned your IP address is dead for whatever reason.
According to the DHCP spec (RFC2131 is the main one) you will then go into a retry sequence at progressively shorter intervals, continuing to try this same server. If at the end of this 'renew' interval you still can't get your lease renewed, you transition to the 'rebind' state of the DHCP state machine and issue a broadcast rebind message. In that case you MAY get a new IP address assigned (from a new dhcp server), depending on how "planned and coordinated" the outage of the original provisioning server was.
As for the need for all users to have an IP address fingerprint, so to speak, I don't quite see how that's possible given that IP addressing is hierarchical. If you physically move, your're on a different subnet or maybe even an entirely different network. The network has no knowledge of where you were or where you're going to, or even a protocol to convey this, so you'd be "non-routable" at your new location. Somehow I think I missed your point because I'm certain you know this. Maybe you weren't referring to the mobile aspect?
As for address shortages, and also tying in to the IP fingerprint concept, what about all the universal non-externally-routable addresses (eg 192.168.x.x) that many people have as "their" IP address, yet many other folks have the same address as "their" IP address? And what about CIDR? How does that fit in with the IP shortage concept?
No doubt practically every subnet in existence is wasteful of IP address space, and no doubt IP addresses are a limited resource since they are from a countable set. But, there already exist many pockets of replicated addresses so the current system as it stands makes effective use of "aliasing." How would we undo this in one fell swoop? And how would we convert a hierarchical addressing scheme into one which had a globally unique identifier attached to it? I suppose there are working groups investingating this very issue, and it's probably a lot more complicated than it appears to the naked eye.
gotta run. |
