Jun-ichiro itojun Hagino wrote:
>
> >> 2.1 RFC1981 - Path MTU Discovery for IP Version 6
> >>
> >> > If Path MTU Discovery is not implemented, then the uplink packet
> >> > size MUST be limited to 1280 octets (standard limit in [IPv6]).
> >> > However, the cellular host MUST be able to receive packets with size
> >> > up to the link MTU.
> >>
> >> The text (3rd and 4th line) is unclear (is it after the reassembly
> >> or before? or am I badly confused?), and may conflict with RFC2460
> >> depending on the link MTU for the cellular host. ???
> >
> >This is before reassembly. Let's try to clarify this with an example. Consider
> >the following situation:
> >
> > link MTU=1500 path MTU=1500
> >cellular host A <--------------> router <--- ... network ... ---> host B
> >
> >
> >The cellular host A communicates with host B. Host A does not do Path MTU
> >discovery, but host B does. Host A will have to limit its packet size to 1280
> >bytes, while host B finds a path MTU of 1500 and will send packets up to that
> >size. Maybe its a bit unclear, since we're stating the obvious that a host must
> >be able to receive packets with size up to its link MTU.
>
> yes, i fully understand the situation.
> how about adding "(before reassembly)" at the end of the paragraph,
> just for clarity?
Agree, that will be more clear, we'll add that.
>
> >> 2.10 RFC2710 - Multicast Listener Discovery (MLD) for IPv6
> >>
> >> (MLD6 is SHOULD) what happens if a cellular node doesn't emit MLD?
> >> will the upstream router always flood multicast datagrams to cellular
> >> hosts?
> > Internet
> > |
> > |
> > |
> > ___________|__________
> > | |
> > | Upstream router |
> > |______________________|
> > | | ....... |
> > | | |
> > | | |
> > host 1 host 2 host n
> (snip)
> >For such a topology, the advantages of using multicasting (including MLD) might
> >be limited. A multicast datagram arriving at the router will have to be sent on
> >each individual point-to-point link to each cellular host subscribed to that
> >particular multicast group. Only the load on the link from the router to the
> >Internet can be reduced by multicasting. This at cost of extra load on the
> >(bandwidth limited) wireless links in the form of MLD signaling.
>
> assuming that we do need to do multicasts (not just link-local):
>
> if you use MLD between "upstream router" and cellular hosts, you
> may be able to cut down multicast groups the upstream router would
> need to forward (if none of cellular host[1-n] is listening to the
> group).
>
> if you don't use MLD, the "upstream router" needs to forward multicast
> packets for every groups host[1-n] potentially join in the future.
>
> am i confused?
You're right, if we do need non-link-local multicast, then the upstream router
needs to be notified in some way about which multicast groups each cellular host
is listening to- Otherwise the upstream router doesn't know to which host it has
to forward specific multicast datagrams and the only option that remains is to
send each multicast datagram to each host, wasting a lot of bandwidth. MLD
between the hosts and the upstream router is a possible mechanism for notifying
the router about the multicast groups the different hosts are interested in.
>
> itojun
Gerben Kuijpers
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
