Hello,
James Kempf wrote:
It seems to me that one of the problems might be that MANET defines a routing architecture which is not based on subnets.
Right.
If a MANET is in isolation, this isn't so much of a problem, because the MANET routing protocol takes care of forwarding.
Right.
However, autoconf is proposing to connect two differently architected networks, the IP subnet routed style and a MANET.
Yes, but this important issue is still subordinate to the main point of getting an address assignment mechanism. And if [autoconf] only did the isolated address assignment, it would still be useful, although not as useful.
So the questions I understand Thomas to be asking (coming from the Internet side) are what is the "gearing" for the "transmission" between these two based on IP subnet routing,
I didn't hear that question, but I can see how Thomas's issues could be beneficially interpreted that way.
while the responses I hear Charlie giving are that there is some empirical evidence that certain kinds of "gearing" work well and they would like standardize those.
I think the main point is that the Internet is supposed to be shielded from the internal structure of the ad hoc network, so that the gearing is more like a disengagement of the gearbox. But that is a very general characterization, with many particular refinements and intermediate design points. I think you would find general agreement on the idea that a gateway should absolutely minimize the amount of information about the ad hoc network that it advertises to the Internet.
Have I got that right?
Sounds about right to me. Regards, Charlie P. _______________________________________________ Int-area mailing list [email protected] https://www1.ietf.org/mailman/listinfo/int-area
