Re: [DNA] Some questions about Simple DNA

[Suresh Krishnan <suresh.krishnan@ericsson.com>]

Hi Bernard,
   Thanks for the comments. Please see responses inline.

Bernard Aboba wrote:
> I was looking over the Simple DNA draft, and had some questions:
> 
> Section 3.3 says:
> 
>    All Router Solicitations and unicast Neighbour Solicitations sent for
>    DNA purposes while addresses are in optimistic state SHOULD include
>    the Tentative Option [4].
> 
> I do not believe that DNAv6 implementations should be required to support
> Optimistic DAD. Prior to completion of DNAv6, a candidate address can be
> considered Deprecated; after successful completion, it can be promoted to
> Preferred. 
> 
> Why should the Tentative Option be set in a Unicast Neighbor Solicitation?
> After all, this is sent to the unicast IPv6 and MAC address of the router, 
> not to multicast IPv6/MAC addresses.  Therefore, there is no risk of
> cache pollution.  Therefore why not establish the neighbor cache entry
> immediately so that incoming packets can be received?  To facilitate
> this, the unicast NS packet should be sourced from the candidate address.

The Router needs to be able to respond to this unicast NS packet. For 
this it requires the L2 address of the soliciting node. If the node 
includes an SLLAO, it might cause cache pollution if there is another 
node who is currently using the same address on the network. This is why 
we need the TSLLAO.

> 
> 
> Section 3.4 says:
> 
>    The Router Solicitation is sent to the All-routers
>    multicast address containing one of the host's optimistic unicast
>    source address [2][5].  If the host is in possession of more than one
>    valid IPv6 address, it MUST send only one router solicitation using
>    any one of its valid IPv6 addresses as the source address.
> 
> Why is one of the candidate addresses chosen as the source?  Since the router 
> MAY respond to this address with a unicast RA, doesn't this risk loss of the
> RA if the candidate address chosen isn't appropriate for the network?  
> Wouldn't it make more sense for the source to be the unspecified address?
> Also, I'd suggest that the RS *not* contain a SLLAO, so as to avoid
> cache pollution. 

If we send out an RS with the unspecified source address, this will lead 
to the router sending a multicast RA to the all-nodes address. This 
might have an undesirable side effect of waking up dormant nodes. I do 
not have strong feelings about this, but others might.

Thanks
Suresh