[09:37:45] --- dthaler has joined
[09:38:35] --- narten has joined
[09:38:46] <dthaler> test
[09:38:57] <narten> I see you!
[09:40:05] --- thomasC has joined
[09:40:11] <thomasC> howdy
[09:43:42] <narten> http://www.ietf.org/mail-archive/web/autoconf/current/msg00924.html
[09:50:27] <dthaler> this doc is a manet arch, not just an autoconf arch doc
[09:51:07] <dthaler> RFC 2461: 
[09:51:26] <dthaler> "Note that all link types (including NBMA) are
expected to provide multicast service for IP (e.g.,
using multicast servers), but it is an issue for
further study whether ND should use such facilities
or an alternate mechanism that provides the
equivalent ND services."
[09:57:42] <dthaler> The quote from 2461 at the top of the link types list is: "Different link layers have different properties.  The ones of concern to Neighbor Discovery are:" (implying there can be other link types than the 3 listed)
[10:07:52] <dthaler> it IS under a single prefix, if that prefix is fe80::/64
[10:07:58] <dthaler> you have no choice
[10:08:00] <dthaler> for link-locals
[10:12:03] <thomasC> Consider 3 nodes, each with *ONE* wireless broadcast interface, and where the reachability is thus: A----B----C
[10:12:13] <thomasC> A and B and C, are they on the same link, Thomas?
[10:13:16] <dthaler> if A and B can communicate without IP forwarding (e.g. TTL decrement), yes.
[10:13:18] <dthaler> can they?
[10:13:26] <narten> depends on how you define things... these are choices that the WG needs to say something about.
[10:13:26] <thomasC> A and B, yes.
[10:13:30] <thomasC> B and C can too.
[10:13:38] <thomasC> But A and C can not, need B to forward.
[10:13:53] <dthaler> so does B forward at L2 or L3 thomasC?
[10:14:03] <thomasC> L3
[10:14:07] <thomasC> IP routing
[10:14:11] <dthaler> then no they're not on the same link
[10:14:31] <thomasC> Well, ok. So A and B are on the same link, B and C are on the same link.
[10:14:34] <dthaler> well let me take that back one step
[10:14:36] <thomasC> But A and C are not.
[10:14:52] <dthaler> because there are still cases where it might be on the same link
[10:15:08] <thomasC> hmm...
[10:15:11] <dthaler> compare to the "off-link model" on Ethernet
[10:15:39] <dthaler> but off-link doesn't apply to link-layer, so can rephrase as
[10:15:42] <thomasC> Well....
[10:15:58] <dthaler> can A send a packet to C's link-layer IPv6 address?
[10:17:01] <dthaler> oops let me fix wording
[10:17:16] <dthaler> offlink doesn't apply to link-LOCAL addresses.  can A send a packet to C's link-LOCAL ipv6 address?
[10:17:31] <thomasC> I'd say no.
[10:17:41] <dthaler> ok then no A and C are not on the same link
[10:22:46] <dthaler> both unicast and multicast, not just multicast
[10:23:13] <dthaler> (and broadcast if you consider IPv4)
[10:23:35] <thomasC> right, yes.
[10:26:05] <dthaler> what it looks like to me is that the arch is trying to define a link model over which the classic IP model at L3 does not hold.  As a result, it tries to restrict the nodes on that link to be only routers that run a constrained set of applications (like routing protocols) that can explicitly deal with the non-IP-model.
[10:26:50] <thomasC> That is the intent, yes.
[10:26:59] <dthaler> and to nodes not on the link (e.g. ones that send traffic that may go over the manet link as transit) it appears as if it was a classic NBMA link
[10:27:21] <thomasC> I.e.....
[10:27:48] <thomasC> {hosts-on-classic-link}-A------B------B-{hosts-on-classic-link}
[10:27:59] <thomasC> Whoops, the last B should be a C
[10:28:34] <thomasC> So A-B-C is, indeed, looking like any other IP-network for the stuff on either side.
[10:28:51] <dthaler> but inside A-B-C it's not the IP model.
[10:29:15] <thomasC> nope, it's MANET interfaces with unusual characteristics running unusual (MANET-aware) protocols.
[10:34:07] <dthaler> yes
[10:34:13] <dthaler> it's not at all clear from Figure 6
[10:38:28] <thomasC> ok, well that is something that we can (relatively) easily resolve, then.
[10:51:01] <dthaler> I would suggest defining a link type of "Semi-Broadcast Multi-Access (SBMA)" or something like that, and defining it as "A multicast link with asymmetric reachability", where multicast and asymmetric reachability are as defined in 2461 (can copy the definitions).
[11:13:43] --- dthaler has left
[11:34:31] --- narten has left
[11:55:19] --- thomasC has left