Re: GSE

[Iljitsch van Beijnum <iljitsch@muada.com>]

On Mon, 17 Feb 2003, J. Noel Chiappa wrote:

>     > Unfortunately, this is hard to get right unless you have access
>     > to the full transport layer state.

> ??? Why do you need the full transport layer state?

You want to switch to another path when the current one becomes
available, but there is no easy way to determine when this happens.
Reliable transport protocols know whether the other end is still
reachable so they can initiate a failover event if this is no longer the
case. A box in the middle would have to monitor the session state. Not
impossible, but not something you'd want to do in a border router
either.

The alternative is sending keepalives to every network you're
communicating with. (This is what MHAP does.)

> Which the end-host does have access to, of course - provided you're doing
> this in the end-host, and not at some border box - another thing GSE alluded
> to, but which wasn't fully worked out, IIRC.

> I don't know, I get a little nervous when functionality is spread out across
> multiple boxes like that; the notion that the binding from identity to
> location is something taken care of by your border router makes me uneasy.

There are good reasons to do it in the hosts and there are good reasons
to do it closer to the edge of the network. Ideally, this choice should
be the 1user's. What we don't want is changes to both router _and_ hosts
for a multihoming solution.