On Thu, Apr 15, 2004 at 01:32:56PM +0200, Tom Ivar Helbekkmo wrote:
> I guess "Mobile IP" is the solution I really want, though, since it
> would fix things like ssh at the same time.  :-)

I still think that Coda is a step in the right direction, to make
the functionality independent of the [presence of] communication channel.
It can (and probably should) be extended to behave well when other
properties of the connection change, besides presence/absence.

Currently we can rather easily support transitions
like IP-conn.A - disconnection - IP-conn.B - disconnection - IP-conn.X ...

Going one step further we could do
 IP-conn - disconnection - non-IP-conn - disconnection - and so on

And of course we could try to keep the state over connection change
as much as it is feasible, i.e. instead of "retry-desperateTimeout-drop"
use like
"retryviaAX-retryviaBX-...
 ...-reevaluateAvailableChannels(say,network interfaces)-retryviaCX-...
 ...-reevaluateServerAddress-retryviaCY-retryviaDY-...
 ...-desperateTimeout-drop".

It is pretty orthogonal to adding different kinds of transport.

For transport we could use anything that delivers rpc2 semantics,
independently of the underlying technologies.

With all of the above, we would be in better position than "Mobile IP"
as we would not be limited to IP :)

My 2c
--
Ivan
P.S. I think "screen" and a reconnecting ssh-wrapper with authentication
agent can achieve "connection persistency"-alike for terminal sessions...