"newbie" <huntedsole@yahoo.com> wrote in message
news:1145121133.688245.145890@v46g2000cwv.googlegr oups.com...
> Hi,
>
> Here I have one question, As I read that for multicast packet
> forwarding, router uses its unicast spanning tree of router by looking
> into its routing table. So while maintaining info about multicast
> groups, does router send query to neighbour routers also??

This depends on the multicast router - router protocol used. Most common
deployed seems to be PIM sparse mode....

this builds trees on actual traffic when destinations ask to be added to the
multicast address group.
>
> Like as we know, for querying about multicast group subscription,
> router will send packet to 224.0.0.1(means to all host in the subnet).
> Dose this "all host" includes its adjacent or neighbour routers(which
> are directly connected with it) also. So that based on their response
> it could maintain routing table.

The subscriptions happen on a per subnet basis - this is really what hosts
do to tell the routers which groups they want to listen to.
Separate protocols handle router to router comms.
>
> So if each router(say A) could receive the neighbour router's(say B)
> multicast group query, then it could inform that router about its
> multicast group subscriptions. So that later, whenever B gets any
> packet to forward, it will fwded to A, only if A has
> any subscription for that group, otherwise not. This will avoid
> unnecessary flooding of multicast packets. Thats why I wondered, if B
> sends a multicast query packet at 224.0.0.1, then will it be received
> and processed by its neighbour router A also??

the local routers elect an IGMP gateway so that only 1 is working at a time
for host queries.
>
> As well as I am not clear with TTL significance in multicast routing.
> Somewhere I read, "For multicast messages(i.e. expect query/report) the
> value of TTL should be greater than one". Dose it mean, by putting any
> TTL value greater than one(say 2) will assure the delivery of this
> packet to all of its group subscribers?? However I don't think so, in
> that case how we decide about TTL value for a muticast packet on a
> originating source.

TTL can be used to limit how far a multicast goes across a network - but
actually using this mechanism is difficult as soon as you have a significant
number of routers and / or a complicated topology.

There are some additional protocols like MSDP that handle joining multicast
domains together - these allow you to limit the scope of a specific set of
multicast addresses.

> Thanks in advance.
>
>
> Albert Manfredi wrote:
> > <myself_rajat@yahoo.com> wrote:
> >
> > > ya I did check with IGMP version(using sniffer), its V3. So host is
> > > sending multicast reports at 224.0.0.22. As I found this special class
> > > D address is ment for IGMP, does this mean that its specifying the
> > > address of local IGMP enabled router.
> >
> > The address 224.0.0.22 goes to all multicast-enabled (IGMPv3) routers
> > within one hop, meaning in the same IP subnet of the host. But only one
> > of these routers becomes the "querier," as the RFCs explain. So IGMP
> > never worries about the specific address of any one router. The routers,
> > if multiple exist in the subnet, negotiate among themselves who gets to
> > be in control, and the hosts can remain as dumb as possible. Always do
> > the same thing, no matter what the router situation is.
> >
> > > And what will happen if the local subnet it having no multicast
> > > enabled
> > > router but simple unicast-router, then how that router will behave
> > > after getting multicast report destined to 224.0.0.22??
> >
> > You would get no IGMP queries from any router, and IGMP reports would be
> > dropped. You will note, for example, that Windows Media Player does not
> > depend on IP Multicast. It can also use unicast, TCP, HTTP, etc.
> >
> > Buy the way, all of this discussion has been about edge multicast router
> > to multicast hosts. There is a whole other discussion about building
> > multicast trees between routers. There have been many schemes developed
> > to achieve that, but it looks like the ones in favor now are under the
> > Protocol Independent Multicast (PIM) banner.
> >
> > Bert
--
Regards

stephen_hope@xyzworld.com - replace xyz with ntl