*>
*> My point remains, a globally meaningful address is something that
*> should only be applied when it is useful for that endpoint to
*> be globally addressable.
*>
That sounds like an appealing statement, but it hides the potential
cost of giving up generality. Back when TCP/IP was young, the
operating systems researchers in Computer Science departments had just
found a new playpen -- distributed operating systems. They disdained
TCP/IP, choosing to implement their OS mechanisms on "bare" Ethernets.
Their statement was that a globally meaningful protocol is something
that should be applied only when it is useful for the endpoints to be
globally reachable. Since all their boxes were local, and for
efficiency, they insisted on running directly over the link layer.
(And BTW, there was only one link layer, Ethernet ;-)) We told them
that the day would come when they would want the general connectivity
of IP, but they were, as I said, disdainful. It took a few years for
them the realize the error of their approach, but they did eventually.
So there is a trade-off here. In general, I think one can say that the
Internet has benefited hugely in the past from taking the approach of
maximum flexibility whenever feasible.
Bob Braden