Re: [DNA-BOF] draft-moore-optimistic-dad-03.txt

[Greg Daley <greg.daley@eng.monash.edu.au>]

Hi Sharkey,

I've got some comments. Please be aware that they'll
have my own personal spin on them (not related explicitly
to DNA-BOF).

The comments follow.

Nick 'Sharkey' Moore wrote:
> G'day all,
> 
> 	well, I've finally gotten around to updating the
> Optimistic DAD draft to -03.  It's really just a few tweaks to
> the rules, but I've also added a mini problem statement just
> to clarify what I think the problems of standard DAD are.
> 
> 	It should, nits willing, appear on the archives soon,
> but in the meantime you can grab it fresh from: 
> <http://bluesky.zoic.org/sharkey/draft-moore-ipv6-optimistic-dad-03.txt>
> 
> 	Let me know what you think -- it's only 13 honest pages,
> so it shouldn't take too long!  I'd really like to get
> an informed discussion going about this at IETF 58.
> 
> -----Nick
> 

I think it is largely good.

I'm unsure though about the usage
of unsolicited NAs (particularly with All-nodes'
multicast destinations).  This could make it hard to
scale the network to contain many nodes, especially
since we're considering opti-DAD on link-local on
each L2 cell change (for every wireless host).


These are the only nit points I have as yet:

The last paragraph of 4 (.0)

indicates that an ON will send out
"a Neighbour
    Advertisement (with Override set to 0) for the address. This NA
    allows communication with neighbours to begin immediately."

I'm not sure if this conflicts with the 'may'
from the section:

"The ON may choose to send unsolicited NAs to the All Nodes Multicast,
   to the All Routers Multicast, or Unicast to the source of the RA
   which alerted it to this new prefix.  This allows flexibility with
   regard to Layer 2 multicast characteristics."

Toward the bottom of page 5 (Optimistic Behaviours).  As it is though,
I'm not sure if the MN is typically aware of the effect that
this form of multicast can have in a heterogeneous LAN.

para 3 section 4.1:

" The Optimistic Node already has the link-layer address of the router
   (from the RA), and the router now has the link-layer address of the
   Optimistic Node (or at least, can find it through standard NUD).
   Communications can begin immediately."

Should NUD be ND?

Section 4.1
the last paragraph seems to be about sending NA O=1.

Is this necessary all the time?

If we've succeeded, there are either no other NC entries than our own
or there's been an attempt by another node to configure the same address
during the same period (which we've beaten by NA defending).

If we've seen DAD attempts, then certainly overwrite with O=1.

Otherwise,  with our O=0 entry, this may be overwritten in
the next N seconds by another ON or Tentative node, but any
interested node will perform NUD soon anyway, and get our updated NA.
Before this time (but after we're no longer tentative), any
attempts to DAD our address will be trumped by our O=1 defence.

I think this removes most of the reason to send unsolicited NA,
even if we've got sessions going.

Please tell me if I've overlooked an important issue here.

I'd really like to get rid of any reliance on unsolicited NA, if
possible.

The rest is checked, and technically OK AFAIK.

Greg