mobility (and multicast...)

[Jon Crowcroft <[e-mail]>]

there was this really cute idea way back in the mid 80s from
cheriton and deering that you could use multicast groups of one and
multicast routing to achieve mobile hosts, which leads me to respond
to yakov's posting of the mobile document to the tuba list recently
with the following...

VAMPIRE = Vicinity Addressing, Mobility Path, Intedrnet Routing Extension

unlike the cell based beakon based approach a lot of mobile IP 9or
isogram) approach has taken, this is based on 2 assumptions.

1. Vicinity Based Addresses - an address i nthe future can be assigned
with hierarchical georgraphic centered assignment, and we can direct
a cast to roughly the right place (a la landmark routing and other
similar approaches). If we have vicinity based multicast (i.e. remove
the TTL type scoping, and put it in addresses as it should be, we can
address groups in a vicinity - a group could be a mobile host (groupo
of one) or mobile LAN)

2. mobility paths: mobile hosts _know_ wheere they are going, so new
schemes like PIP permit us to efficiently source  route a packet to
the vicinity, then track the local moves....a mobile host or
designated mobile router on a mobile lan could use scoped multicast
to advertise its moving vicinity-ness...

3. i believe mobility should be _integral_ in routers and not involve
more processes/servers etc....if hierarchical multicast addressing and
routeing are in place, i think something could evolve that is very
clean....

anyhow, sorry to cross post, just thinking out loud...

 jon