Layer 2 MAC-layer group management [Was: New draft on embedding the RP address in IPv6 multicast address]

[Bill Nickless <nickless@mcs.anl.gov>]

At 02:48 PM 10/17/2002 -0400, Brian Haberman wrote:
> Sounds to me like you just described a MAC-layer group management
> protocol.  Since the IETF generally stays at Layer 3 (Sub-IP
> excluded), it is out of our control.
Exactly right.  So why are the IETF working groups like MAGMA trying to 
specify Layer 2 switch multicast behavior?

> Perhaps you should petition the IEEE to develop this protocol or
> write a draft on how a layer 3 devices should advertise their
> interest in receiving packets on layer 2 addresses.
The IEEE 802.1p committee did develop a MAC-layer receiver interest 
protocol.  See the work on GARP/GMRP.  A good introduction to that work can 
be found in the second section of 
http://www.ieee802.org/1/mirror/8021/tobagi/garp-gmrp-timers.pdf .

I did write a draft, which included a discussion of how layer 3 devices 
should advertise their interest in receiving packets on layer 2 
addresses.  See the first part of section 22 (page 10) of

ftp://ftp.ietf.org/internet-drafts/draft-ietf-mboned-iesg-gap-analysis-00.txt

It's written to first fix the truly unpatchable problem at multicast 
exchange points, where not all exchange point participants have peering 
agreements--leading to situations where datagrams might need to traverse 
the exchange point media more than once.  That happens all the time for 
unicast datagrams, but can't happen with multicast datagrams.  (Ibid, 
Section 20, page 8.)

Then, once the technology is tested and proven in the limited cases, 
leverage it at the network edge for large-ish segments (like at Berkeley).

===
Bill Nickless    http://www.mcs.anl.gov/people/nickless      +1 630 252 7390
PGP:0E 0F 16 80 C5 B1 69 52 E1 44 1A A5 0E 1B 74 F7     nickless@mcs.anl.gov