Christian,
IMHO, every identifier ends up being routed, at least in some context.
I think I fully agree with you, but IMHO we need to be very careful
about the terminology here. For example, I can easily imagine an
overlay network, running on the top of the current IP network, using
DHTs to "route" flat identifiers. If we want to call that kind of
"application-layer" forwarding routing, then I fully agree.
For example, there is a good case that on a disconnected ad hoc
network,
it makes more sense to use the identifiers you have than to create
some
new addresses. Indeed, if one is willing to have individual host
entries
in a routing table, then one can use any identifier.
I concur. But as you well know, that does not scale. Once your ad
hoc network grows to a size where clustering becomes a necessity, you
can no longer use fixed flat identifiers for routing purposes. [Note
that I am not saying that you need hierarchically organised
addresses; more desirable would probably be dynamically assigned
locators that encode clustering information in some better-than-
hierarchically-assigned way.]
--Pekka
_______________________________________________
Int-area mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/int-area
