[Igmp-mld-bis] Fwd: [pim] Volunteers needed for work on progressing IGMP/MLD on the standards track

Forwarding this informative email.  Do we want to create issues in the GitHub repo to capture some of these?

Regards,
Femi

> Begin forwarded message:
> 
> From: Linus Lüssing <linus.luessing@c0d3.blue>
> Subject: Re: [pim] Volunteers needed for work on progressing IGMP/MLD on the standards track
> Date: 13 December 2018 at 14:32:29 GMT
> To: Stig Venaas <stig@venaas.com>
> Cc: mcast-wifi@ietf.org, pim@ietf.org
> 
> And a five more inconveniences that came to my mind:
> 
> 
> # IGMP Query 0.0.0.0 source address
> 
> As snooping switches have an interest in learning about multicast
> listeneres, they often implement the querier protocol. However,
> they might not implement DHCP for instance, leaving them with no
> assigned IPv4 address to use for IGMP queries.
> 
> In that case, such snooping switches will usually use the 0.0.0.0
> source address.
> 
> The issue with that however is that according to IGMP they would
> always win the querier election as it is the lowest address
> possible.
> 
> This is undesirable due to the following two reasons: For one
> thing a snooping switch might not always be in a favorable
> position topology wise. A real multicast router for instance is
> probably in a better spot but would always loose the election. For
> another, when multiple such snooping switches use the 0.0.0.0
> address then this creates quite some turbulences: The selected
> querier will change frequently.
> 
> If I remember correctly, there was some debate on the Linux bridge
> mailinglist in the past regarding how to treat an IGMP query with
> a zero-source address.
> 
> For reports, there is a nice section regarding the source address
> (RFC3376, section 4.2.13. "IP Source Addresses for Reports").
> However there is no equivilant for that for queries.
> 
> Ideally, 0.0.0.0 would be excluded and ignored from the querier
> election mechanism, I think.
> 
> 
> # MLD Query :: source address
> 
> RFC4541, section 3 ("IPv6 considerations") says:
> 
> "[...] the switch should use the null IP source address (::) when
> sending said queries"
> 
> While RFC3810, section 5.1.14 ("Source Addresses for Queries") says:
> 
> "All MLDv2 Queries MUST be sent with a valid IPv6 link-local
> source address. If a node (router or host) receives a Query
> message with the IPv6 Source Address set to the unspecified address (::), or
> any other address that is not a valid IPv6 link-local address,
> it MUST silently discard the message and SHOULD log a warning."
> 
> Currently, the Linux bridge for instance adhears to the latter, ignoring
> all MLD queries with a :: source address.
> 
> 
> # Query type for Querier Election
> 
> RFC3376, 6.6.2. Querier Election:
> 
> "When a router receives a query with a lower IP address, it sets the
> Other-Querier-Present timer to Other Querier Present Interval and ceases
> to send queries on the network if it was the previously elected
> querier."
> 
> The "receives a query" should probably be a "receives a general
> query". Similarly for MLDv2 (RFC3810, section 7.6.2).
> 
> At least that's what the Linux bridge currently does. It only
> processes general queries for the querier election and for
> multicast router port additions.
> 
> 
> # IPv6 link-local multicast to multicast routers
> 
> Currently it seems slightly unclear whether data with an IPv6
> link-local destination needs to be forwarded to multicast routers.
> 
> Ideally, this would not be necessary as link-local multicast by
> definition will not be routed. This would potentially save quite
> some traffic as in a large broadcast domain there are quite a few
> such messages for IPv6 Neighbor Discovery, for instance.
> 
> However, this is probably, unfortunately not do-able right now.
> At least when  there are IGMPv1/2 / MLDv1 hosts present. Due to
> the report suppression link-local multicast will need to be
> forwarded to the selected querier / some common host which
> received all reports, to avoid snooping switches introducing
> multicast packet loss.
> 
> RFC4541 does not seem to make any special recommendations
> regarding IPv6 link-local multicast.
> 
> ..o(One thought that crossed my mind was, whether it might make sense
> to add a flag to Multicast Router Advertisements which if set
> explicitly indicates that the according router is not interested
> link-local multicast.)
> 
> 
> # Querier => Multicast Router implication?
> 
> Ideally, RFC4286 ("Multicast Router Discovery") would become the
> sole source to detect multicast routers. And RFC4541 would not
> (need to) recommend to add queriers to the multicast router list
> anymore. Especially as snooping switches often implement IGMP/MLD
> querying, too, this is a little inconvenient.
> 
> However, since RFC4286 is neither mandatory nor widely
> implemented and the report suppression issue still present this
> too is probably, unfortunately more a wish-list item for the
> distant future.
> 
> 
> Regards, Linus
> 
> _______________________________________________
> pim mailing list
> pim@ietf.org
> https://www.ietf.org/mailman/listinfo/pim