[Cu-wireless] Re: phase iii network: v4/v6 addressing & Internet connectivity

David Young dyoung at pobox.com
Fri Mar 12 17:31:56 CST 2004


On Fri, Mar 12, 2004 at 08:37:12AM -0600, Joe Pickert wrote:
> The first thing that springs to mind is mobile ip. I am pretty sure there
> is at least one open source implementation available. If you consider the
> gateway that the connection is currently established on as the "home
> address", and the gateway with better metrics that you are transferring to
> the "attachment point", then mobile ip seems to fit the bill. Granted, you
> still have a less than optimal path for the connection overall (since
> mobile ip is basically just redirection). However, at least more of the
> traffic can be through paths off-cloud, which are presumably better, and
> all on-cloud traffic can be using the best available paths. You also have
> the nice benefit of all connections being totally mobile within the cloud.
> Of course this assumes your software in managing the cloud side of all
> gateways. Is this an incorrect assumption (or perhaps an undesirable one)?

Mobile IP is something for us to look at, but I think that for the amount
of effort it would take to deploy, it might not add a lot of value.
I am concerned about the overhead of "triangle routing."  David Chiluk,
one of Ralph Johnson's students, astutely observed that on the assymetric
net connections that we can count on, triangle routing may not only be
wasteful, but practically unusable!

[BTW, I think that with Mobile IPv6, triangle routing can be avoided.
I think that it has to do w/ identifying TCP sessions using the (EUI64,
port, EUI64, port) tuple instead of the (IPv6, port, IPv6, port) tuple.]

Dave

-- 
David Young             OJC Technologies
dyoung at ojctech.com      Urbana, IL * (217) 278-3933



More information about the CU-Wireless mailing list