[prev in list] [next in list] [prev in thread] [next in thread] 

List:       namedroppers
Subject:    IPv6 design flaw to force all-nodes multicast
From:       Masataka Ohta <mohta () necom830 ! hpcl ! titech ! ac ! jp>
Date:       2007-01-12 2:15:13
Message-ID: 45A6EF31.9040300 () necom830 ! hpcl ! titech ! ac ! jp
[Download RAW message or body]

Pars Mutaf wrote:

>>Who cares what happens at L2? That is not a concern to IP.

> But we can build new applications comfortably if we know that the
> signaling and energy costs were minimized ?
> (I mean regardless of L2 specific details)  

With the obvious example of you,

            link-local subnet prefix | 64bithash("johnsmith.local")

supporting a L3 multicast address over broadcast L1 is just as
costly as supporting a L3 unicast address over broadcast L1,
regardless of L2 specific details.

Wireless L2 protocols not supporting L3 multicast is just poor.

In addition, wireless L2 protocols can also naturally support
L2 broadcast. If power consumption of terminals but not central
stations is the issue, broadcast can be issued only from (or
relayed by) central stations and can be rate limited.

Thus, all the properly designed L3 or upper protocols can enjoy
properly designed L2 multicast at any rate and L2 broadcast maybe
with rate limitation. It has nothing specific to mDNS.

However, there is an IPv6 design flaw that all the nodess join
all-nodes multicast, which is just a common multicast address
for L2. The flaw was derived from a misconception that multicast
were less costly than broadcast, though broadcast is less costly
than multicast.

All-nodes multicast can cost a lot if someone use it frequently.

For example, ND is specified with:

   o  MinRtrAdvInterval 0.03 seconds

That is, as IPv6 is not propderly designed, L2 under it MAY
suppress multicast entirely and SHOULD supress the all-nodes
multicast address.

A dirty workaround is to supress rate of L2 address of all-nodes
multicast and never use the same L2 address for other rate-unlimited
multicast addresses.

						Masataka Ohta


--
to unsubscribe send a message to namedroppers-request@ops.ietf.org with
the word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/namedroppers/>
[prev in list] [next in list] [prev in thread] [next in thread]