[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
(usagi-users 02921) Re: IPv6 Multicast on recent Linux Kernels
- To: usagi-users@xxxxxxxxxxxxxx
- Subject: (usagi-users 02921) Re: IPv6 Multicast on recent Linux Kernels
- From: David Stevens <dlstevens@xxxxxxxxxx>
- Date: Tue, 20 Apr 2004 23:51:53 -0700
- In-reply-to: <20040421.144640.96398854.yoshfuji@linux-ipv6.org>
- Reply-to: usagi-users@xxxxxxxxxxxxxx
"Hideaki YOSHIFUJI" <yoshfuji@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote on
04/20/2004 10:46:40 PM:
> Multicast router SHOULD be required to "forward" multicast packets.
This question came up in another context for me a couple months ago
and if I interpreted it correctly, and remember correctly, one of the
MIPv6 RFC's says the following (paraphrased).
If an MN joins multicast groups (including link-local groups) on a
network and then moves, the home agent should ("SHOULD", not "MUST")
join those groups and forward multicasts from that network to the
MN, at least when the groups aren't available on the new foreign
network.
So, in that case, which was not implemented in MIPL when it came up,
multicast packets would be delivered via the HA to the MN. But that
has nothing to do with whether or not there's a multicast router on
that link. Multicast routers would do forwarding on behalf of the
entire link, and use MLD and routing protocols to determine the
forwarding. For this case (which I think is what the original
question was about), the HA is simply doing host-side MLD on behalf
of the MN and forwarding multicasts for those groups only to the MN.
So, the HA is forwarding multicast packets, but it is not a multicast
router.
If implemented, an MN could participate in multicast groups on
its former network without any multicast router present, the same
way any host that isn't mobile can. It just wasn't implemented in
MIPL yet, and I believe it still isn't.
+-DLS