Hi Itojun,

> i still don't understand why you are trying to impose additional
> restriction for link-local multicast groups (maybe i'm dumb).
> without MLD joins for link-local multicast groups, default routers
> won't be able to know which multicast group the 3GPP node is
> interested in. there's no special text available in RFC2710 for
> p2p links.

First some motivation in the practical situation. The
GGSN has no applications, and hence there wouldn't
be a link-local video distribution from it, as an
example. Furthermore, I believe IPv6 ND packets
are sent to the link regardless of the state of the MLD
situation on e.g. a Solicited Node address -- or perhaps
I have missed some text somewhere? Hence there'd be no use
of the MLD for the IPv6 routing functioanality either.
Obviously, there are no switches around in these interfaces.
Finally, MLD packets do consume bandwidth, particularly if
they need to be sent even when there is no traffic and we
are just listening e.g. on our solicited node mcast address.
(I'm not exactly sure how much this bandwidth is -- perhaps
someone could enlighten me? At least some initial packets
have to be sent, but is there a periodic refresh as well?)

But I do understand that the RFCs must be followed.
However, I wonder if you Brian could say something
about the background for the link-local and the
Solicited Node multicast address requirements,
were they done specifically to allow switches to operate
or for some other reason?

I also wonder about text in Section 2, which says
that  MLD should be on all interfaces from which
an application or upper-layer protocol has requested
reception of multicast packets. I wonder what "upper
layer" means in this particular case...? If it
means above IP then we could argue that we are following
the RFC.

I also wonder if the timers -- which are configurable
per the RFC -- could be tuned to minimize overhead?

Finally, I have a compromise proposal. I do appreciate
Itojun's point some time ago that MLD even on a p2p
link would allow the router to reduce traffic to the
link. This is because it would know whether there are
listeners or not. Perhaps in the future there will be some
new parts of IPv6, or some new applications for which this
would be useful, even in a 3GPP environment. However,
I do wonder about the Solicited Node multicast address
case. Assuming the rules about this are in the RFC
for the support of switches, could we use MLD for
all link-local and global addresses, except for all
nodes and solicited node multicast addresses? Given
that I can't find a place where ND would depend on
MLD, I think this would work well. And the case
is important, as all hosts would have these addresses
and would have to announce them over the network
otherwise. So, the proposal is to have the following text:

   2.9 RFC2710 - Multicast Listener Discovery (MLD) for IPv6
 
   MLD should be supported by cellular hosts.
 
   2.9.1 MLD in 3GPP networks

   Within 3GPP networks, hosts are connected to their
   default routers (GGSN) via point-to-point links.
   This arrangement also precludes any network
   switches. Consequently, hosts cannot communicate
   directly with each other using link-local addresses.
   Furthermore, none of the entities along the link
   will be doing any address-based switching, and
   all messages sent to the point-to-point link
   normally arrive to the other end. As
   Neighbor Discovery operates without relying on
   MLD, joining on solicited node and all nodes
   multicast addresses is not required. 

   However, MLD is required for all other multicast
   addresses in all scopes.

Jari
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to