Re: [DNA-BOF] BoF Last Call on DNA WG Charter.

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

Hi Bernard,

Bernard Aboba wrote:
>>Indeed it may be, if nodes are consistent in
>>getting their addresses from the DHCPv6 server.
>>The current status from RFC-2462 is that hosts
>>DAD their addresses even if obtained from DHCP
>>on their original allocation.
> 
> 
> This is true for DHCPv4 as well, because the check is not required for the
> DHCPv4 server.  The question is whether DAD needs to be
> *rerun* on re-connection.

Indeed.

> 
>>This is not because of problems with DHCP lease
>>state, but because of autoconfiguring nodes (for
>>example, if A=1 in an RA).
>>
>>I'd guess that the same types of autoconfiguring nodes
>>would be the 'troublemakers' for DHC hosts which
>>left a network temporarily and came back with a valid lease.
> 
> 
> I'd agree that if DHCPv6 and autoconfiguration are used at allocate
> addresses *from the same prefix* there will be an issue.  However, this
> seems like a bad idea to me, for a number of reasons.  Is this really a
> case we need to consider?

I think that it's currently possible to have both
O=1 in the RA, but A=1 in a prefix.
Whether or not it's been stated that prefixes
should be advertised with A=0 in a document
somewhere is something else.

It is certainly explicit in RFC-2462 that these
prefixes (A=0) are silently ignored by the
autoconfiguring host.  In this case, I'd try not
to redo DAD (but still may do some form of DAD
originally).

If there is an address advertised as A=1 for which
we have a valid lease, I'd assume the same treatment
for this address as stateless addresses in the same
prefix.

Greg