Re: [pim] Simple join failure notification for PIM-SM multicast routing

[John Zwiebel <jzwiebel at cisco.com>]

2cents:

-- GU messages received by an RP should stop there unless the RP
	has (S,G) state with an OIF-list.  Sending a GU down a
	shared-tree doesn't tell us anything.  If it is important
	to know that a specific source is missing, I'd say that
	the receiver should have been using SSM.  (another point for
	my "dump ASM" campaign.
-- in 2.2 you state that the GUs are sent to the DRs, but its only
	the DR on the fist-hop that this applies to.  Intermediate
	routers may or may not be a DR on any interface.
-- It was suggested that the GU might be useful in getting a downstream
	router to recognize a upstream failure so it might join to
	a different path.  I see danger in this path.  It could interfere
	with the unicast routing protocol that PIM is dependent on.
-- the spec doesn't cover what a host might do with this information,
	if the user doesn't get the report and actually do something
	with it, then I don't see how it is useful.
-- Figure 3: I'd prefer it be called "source address" vs unicast address
-- I don't see the usefulness of the L flag.  If a downstream router  
never
	gets one does that mean anything?
-- R and S bits.  May not be necessary.  Sending S information down the 
	shared tree doesn't mean anything to me.  If the RPF to the
	source == RPF to the RP, receipt of a (*,G) GU means the same thing
	"you can't get there from here".
	ie, a Group Record with 0 source addresses means its an ASM group
	the last-hop router should never have been able to join to the SPT
	so receipt of a GU that tells it about a given source doesn't mean
	anything.  Even in the case where the (S,G) state is set up and then
	the share/SPT goes away at the same time, I think one could do away
	with having to specifically address the shortest path since they
	overlap and the last-hop knows that they overlap.

	in cases where the last-hop RPFs to the same spt/RPT router but
	above that point things diverge, what is the advantage of getting
	S info down the RPT?  Either data stops or it doesn't.  If it
	doesn't, you know the SPT is still up so you don't care about
	that.  But you don't know that the RPT is down, which becomes
	clear when you get the GU with the GR and 0 sources.

	ie, if you do decide to continue with this draft (whether it be in
	ICMP or PIM) as a last-hop I want to know that the (*,G) is down or
	the (S,G) is down.  If the (*,G) and (S,G) overlap to the point of
	failure, there is no additional information in separately specifying
	each (S,G).


	

On Jun 15, 2006, at 1:44 AM, hoerdt at clarinet.u-strasbg.fr wrote:

> Hello,
>
> A new internet-draft is available in the ietf repository :
>
> http://www.ietf.org/internet-drafts/draft-hoerdt-pim-group- 
> unreachable-00.txt
>
> Title :
> "Simple join failure notification for PIM-SM multicast routing"
>
> Abstract:
>
> "Currently, PIM-SM is the most widely deployed multicast routing
> protocol in the Internet. However in IPv6 this multicast protocol
> lacks a simple connectivity debugging tool, making it difficult to
> identify multicast routing problems.  When a host wants to receive a
> multicast group or channel, if the corresponding  PIM-Join message
> cannot be propagated in the network, nothing exists to inform the
> receiver or the network about this failure. We propose a simple error
> notification message based on ICMP/ICMP6 which is able to indicate
> where and why the error happened in the network without using a
> multicast traceroute mechanism."
>
> We would like to present this work in the PIM working group at the
> 66th IETF/Montreal PIM. Before that, we would like to get comments and
> engage discussion with the PIM and MBONED working groups in the
> mailing lists.
>
> thanks,
>
> Hoerdt Mickael
>
> _______________________________________________
> pim mailing list
> pim at ietf.org
> https://www1.ietf.org/mailman/listinfo/pim

_______________________________________________
pim mailing list
pim at ietf.org
https://www1.ietf.org/mailman/listinfo/pim