RE: dual stack & IPv6 on by default

[Mika Liljeberg <mika.liljeberg@welho.com>]

Hi Jim,

On Wed, 2003-03-12 at 05:24, Bound, Jim wrote:
> > So what I'm trying to say is that the rule "If the Default
> > Router List is empty, the sender assumes that the destination 
> > is on-link" doesn't seem to make sense in a multi-homed host. 
> 
> If the node has received no prefix or address knowledge it is a means to
> still try to send the packet.  But in that case as I said above it would
> have to be a default link too or I guess a round robin decision. 

This, I think, is the first time during this discussion that you're
directly addressing the question I am trying to raise.

Default link or round robin? Or try all routerless links at once? That's
my question, and RFC2461 gives no answers at all.

Let's face it, "default interface" is a trivial solution. It might work
in a specific case but not as a general answer. "Round robin" won't work
properly either, since the result is non-deterministic. "Try all
routerless links at once" might work but, hey, that's pretty ugly. There
being no good answer, our current implementationd doesn't try at all.
That, at least, is deterministic.

To take things back in the direction of dual stack hosts and transition,
how does the above fit with the destination address selection algorithm
in RFC3484 and the first rule:

   Rule 1:  Avoid unusable destinations.
   If DB is known to be unreachable or if Source(DB) is undefined, then
   prefer DA.  Similarly, if DA is known to be unreachable or if
   Source(DA) is undefined, then prefer DB.

Is the name resolver implementation supposed to do a route search only,
or is it supposed to try to apply the "no routers -> assume all
destinations are on-link" rule as well?

Of course, RFC3484 doesn't give much guidance either:

      Discussion:  An implementation may know that a particular
      destination is unreachable in several ways.  For example, the
      destination may be reached through a network interface that is
      currently unplugged.  For example, the implementation may retain
      for some period of time information from Neighbor Unreachability
      Detection [14].  In any case, the determination of unreachability
      for the purposes of this rule is implementation-dependent.

An implementation can't really know that a destination address is
unreachable, unless it does a route search and, failing to find one,
proceeds to try to ND all routerless links as well. Is that what we
should do?

	MikaL