IPv6 Neighbor Discovery On-Link Assumption Considered Harmful
draft-ietf-v6ops-onlinkassumption-04

[Ballot comment]
I'm glad to see documentation of design decisions (& changes thereto).  I wish there was more of this, so there could be less "What were they thinking in 2005 when they removed the on-link assumption!?" in 2010 ;-)

[Ballot comment]
I found Section 3.3 very hard to parse.  This section, in particular:

  The second option
  might succeed in reaching a destination (assuming that one is
  reachable) but is more complex to implement, and isn't guaranteed to
  pick the correct destination.  For example, there is still ambiguity
  about which link to use if more than one node answers the
  solicitations on multiple links.

In previous parts of Section 3, we seemed to be dealing with a host attempting to send
a packet using a known AAAA record.  In this section, it's not clear whether the argument
is that a host with multiple interfaces has an interface selection problem that is unresolvable
because of on-link assumptions, or that the on-link assumption again causes a poor ordering
of attempts to reach a host, or causes ambiguity because of non-globally unique addressing.
Some clarification would probably help here.

The working group has to decide whether this document really needs to
published.
In any case it needs more work by the working group before it can be
published.