Le 31/03/2011 17:02, Jong-Hyouk Lee a écrit :
Hi, Alex.
Plz, see inline. I fully failed to understand your words.
On Thu, Mar 31, 2011 at 4:53 PM, Alexandru Petrescu
<alexandru.petre...@gmail.com> wrote:
Le 31/03/2011 16:19, Bob Hinden a écrit :
On Mar 31, 2011, at 1:15 PM, Scott Brim wrote:
On Thu, Mar 31, 2011 at 11:32, Bob
Hinden<bob.hin...@gmail.com> wrote:
They could use a fixed car address and use mobile IPv6 with
a provider based address.
That would help. A rendezvous point (e.g. home agent) always
helps to protect confidentiality. Plus a firewall to protect
the always-out-of-date vehicle software. They can work out
how to do path optimization later.
I agree. If you put the "home agent" in the middle of the
internet (for some value of middle), then path optimization
probably isn't that important.
Makes some sense. This "middle" depends on latency; and core
network has an order of magnitude (sometimes 2) less than even the
first hop to vehicle.
In the context of ITS communication, how do you expect the movement
of vehicles?
Right, mouvement of private vehicles is unpredictable. Some common
sense may reduce the space to, say, mostly a continent. And even some
vehicles like trains have very highly predictable trajectories.
Not only ITS should be considered, but also operators of public
transportation networks (if they're not already part of it).
Then, how do you decide the location of home agent for reducing the
routing performance for user data packets destined to hosts attached
to the router of a vehicle?
The direct seller of vehicle (concessionnaire, garage) may decide to
host HA relativley close (geo terms) to where the vehicle is owned,
suffices HA to be on wire instead of wireless, which is rather natural.
The parts manufacturer too. The transportation operator, etc.
The common wisdom of needing to host a webcache closer in order to
deliver data quicker to http clients assumes the http clients use wire
too (as the servers do). The latency difference between pc's wire and
server's wire is much smaller than between server's wire and wireless.
That means that it may suffice to put the HA on wired link,
anywhere in the Internet, and relatively close (in terms of
hopcount) to where the handover is about to take place.
In the context of this thread, Bob mentioned about the routing
performance, exactly saying user data packets' routing performance,
not saying handover latency mainly consisted of 1) movement
detection at IP level; 2) address configuration; and 3) location
update (BU/BA) to the home agent.
Right, I didn't mean handover performance.
Routing performance... hmm... I believe there is not much routing
performance gain when trying to bring HA as close as possible to the
handover place.
Alex
Cheers.
And not necessarily put it where eg
akamai stuff sits.
Alex
Bob
--------------------------------------------------------------------
IETF IPv6 working group mailing list ipv6@ietf.org Administrative
Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
--------------------------------------------------------------------
IETF IPv6 working group mailing list ipv6@ietf.org Administrative
Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
