On Oct 15, 2009, at 07:18, Pascal Thubert (pthubert) wrote:
As opposed to NDP, this form of NUD is not symmetrical. Traffic from
the
router does not trigger our NR/NC.
Right.
So what kinds of NUD might one need?
Host->Router is indeed best handled by the NR/NC process, and could
also benefit from NUD-like upper-layer indications of liveness.
Router->Host? 4861 NUD is about finding out when a node is gone or
has changed its MAC address.
In a wireless network, nodes are obscured, going away and coming back
all the time.
Using NUD as the carrier for a reachability protocol between routers
and hosts is a distinct possibility, but may simply have the wrong
granularity.
I'd like to understand the requirements for such a reachability
protocol a bit better.
And, again, I see a much stronger link to reachability protocols used
by the routing system than to ND.
(I still believe MAC-layer ACKs are good for quickly handling
*changes* in radio range reachability, even if they don't always
detect all non-radio issues. Oh, should NR/NC maybe be a three-way
handshake?)
Host->Host? Much of the same. This is also about reachability, and
in contrast to Ethernet ND the on-link decision is gated on
reachability as well. And, again, (the lack of) MAC-layer ACKs
provides the quickest possible feedback that there might be a need to
switch back to a router.
But it also would be useful to occasionally get a "end-to-end
confirmation between neighboring IP layers" as in the last sentence of
section 7 of 4861.
(And, back to question 1, there also is a need to find one or more MAC-
layer addresses of the other host -- whatever you use to know it's
there will also reveal at least one.)
My current take is that we are looking for something almost, but not
entirely unlike NS/NA.
(Fixed the subject line; this is about subject 2.)
Gruesse, Carsten
_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan
