Re: [dna] Review of draft-ietf-dna-simple-11

["Laganier, Julien" <[e-mail]>]

Erik Nordmark wrote:
> 
> Bernard Aboba wrote:
> > Erik said:
> >
> > "Perhaps the reference to "STALE" is not clear above?
> > STALE refers to a state in RFC 4861 which implies that the host would
> > trigger NUD sooner rather than later. If it isn't set to STALE then
> > it might take up to 30 seconds more to detect that a router has gone
> > dead. Thus a neighbor cache entry of STALE doesn't have anything to 
> > do with a routing table entry (unless the implementation is broken.)"
> >
> > Since simple DNA involves sending out NS and RS packets, if the host
> > has a valid address on the network corresponding to the STALE entry,
> > it will effectively complete NUD as part of DNA.  Also, by sending
> > out the RS it will discover whether the router whose entry has
> > been marked STALE is actually there or not.
> >
> > Given this, why would a host take 30 seconds to detect that a router
> > has gone dead?  Simple DNA should have completed (with a new default
> > router entry) long before that.
> 
> I was merely responding to Ted's concern, which is unrelated to the use
> of STALE state.
> You are asking a different question about the utility of STALE.
> 
> I have no go over the whole document more carefully, but AFAICT when
> the host has moved to a new link it needs to make sure old neighbor cache
> entries don't get in the way. That isn't specified in section 4.8.
> Marking the NCEs as STALE definitely help with that for the same that
> the same link-local addresses are in use on the old and new links.

IIRC marking old NCEs as STALE takes care of a case where the routers on the old and new link have the same link-local address. By marking the old NCE as STALE, one make sure that the new NCE in REACHABLE state will be preferred over the old one (the new NCE is in REACHABLE state as a result of receiving a solicited RA, or completing NUD with the router.) 

--julien