[e2e] node addresses vs. interface addresses

[J. Noel Chiappa]

    > From: Joe Touch <touch at ISI.EDU>

Now that I've thought this through a little bit more, I came up with a
different way to characterize your scheme. (For both variants: i) use host
routes over a limited scope, and ii) use tunnels.)

    >> The alternative to loading everybody with the problem is to shift the
    >> load to the multihomed host and its peers.

    > Hiding it _does_ shift the load to the multihomed host and its peers.
    > See the paper for details.

These schemes are both geared toward what some call "host multi-homing" -
i.e. a single host has two interfaces - and usually both connections are
within a fairly small scope of the connectivity topology.

Alas, that's not what most multi-homing users want; they want "site
multi-homing", where some organization (typically a company) wants robust
service and so is connected twice to the rest of the Internet fabric; these
connections are usually widely separated in the topology (typically by
signing up with two different ISP's).


The "use host routes over a limited scope" clearly doesn't work there - the
problem is having the distant parties send their packets to the right ISP
to even get to the site - i.e. the same old multi-homing problem.

For the tunnel scheme, the analysis in my previous message holds: unless the
"home station" it itself multi-homed, you haven't really solved the problem.

	Noel