Re: I-D ACTION:draft-hoon-mobileip-undad-00.txt

Hi Jaehoon,

jaehoon.kim wrote:
> Hi, Greg
> 
> Thanks a lot for your interest in our draft.
> 
> I have a question about following sentence.
> 
>>The ping ff02::1 won't work.
>>Existing hosts don't have to respond to the message.
> 
> 
> FF02::1 is already reserved in RFC.

Yes.

> And futhermore, I could not find that some host may not respond to ff02::1
> in any RFC due to firewall.
> Let me know reliable reference.

Page 13, RFC-2463. Third paragraph.

"...An Echo Reply SHOULD be sent in response to an
   Echo Request message sent to an IPv6 multicast address."

SHOULD is not mandatory like MUST (as per RFC 2119).
It is approximately equal to RECOMMENDED.

Firewalling is just one reason not to send the response.

If the response was mandatory, firewalling (for example)
would not be applicable, and the response would always be
received.

I haven't seen any system which requires responses from
all hosts (or all groups of hosts) on a link except for
MLD Queries.

> I usally used ff02::1 or ff02::2 in our lab and ETSI test event.
> I just have been confused in what you mentioned.

The issue is that if a node refuses to respond to
ICMP echo requests, then it will not be in your
list of hosts if you have not received unicast packets
from it recently.

This means you may allocate the address to another, not
knowing that it is already is present on the link.
I think that's what DAD was trying to avoid.

The way your system works now, the router may
incorrectly allocate an address to a new device
(because its state was incomplete). The router
may respond first (This is likely to because the router
response is configured for speed and the host which
owns the address may be powersaving).

In this case, both devices have the address configured
as non-tentative when the DAD defense comes (up to a
second later).

The alternative of keeping the addresses on the configuring
host tentative is already defined in
draft-moore-ipv6-optimistic-dad-03.txt.
This requires no state server.

I'll try to get a draft of MLD-DAD back onto the server
for comparison, so that I can illustrate some of the
issues in state maintenance.

Greg