RE: [DNA-BOF] FW: I-DACTION:draft-park-dna-ipv6dadopt-requirement-00.txt

[Soohong Daniel Park <soohong.park@samsung.com>]

Hi Greg

I really appreciate your valuable feedback...:-)
Please look into my inline response.


> 
> I've been reading your document.
> 
> I know it's a draft, so I'm not worried about
> style unless the meaning is not clear.

Yes, don't worry...

> Overall it's good.
> 
> Normally I'm not sure if hints about
> implementations (section 4) are placed in
> requirements/goals documents.
> The section provides a good non-specific
> background on the state of the art, though, so
> I think it is helpful.

Could you please explain more detail ?

> Section 5 starts with:
> 
> "5. Requirements for IPv6 DAD Optimization
> 
>       This section describes the requirements for IPv6 DAD 
> Optimization
>       and following requirements are referred to as IPv6 DAD 
> Optimization
>       solutions hereafter.  "
> 
> The meaning here is not clear.
> 
> Do you mean something like this?:
> 
> "5. IPv6 DAD Optimization Goals
> 
>     This section defines a set of goals and requirements for IPv6
>     DAD Optimization.  Within this section, future solutions which
>     embody the goals presented here are referred to as
>     'DAD Optimization solutions'. "

correct.

> 
> for RQ2. I'm not sure what is meant by the phrases:
> 
> 
> " For effective DAD optimization, there are no restrictive
>    requirements to modify Neighbor Cache entry in the specific cases.
>    Immediately on completing the appropriate DAD procedure for
>    either original DAD or DAD optimization, all Neighbor's NC
>    (Neighbor Cache) MUST be updated appropriately."
> 
> The first statement regarding the recoverability of the NC 
> modifications is clear though.
> 
> Are you trying to express that:
> 
> "Changes to peers' Neighbor Caches which are due to DAD 
> Optimization must be repaired in the case of a collision. 
> Additionally, cache state created by optimizations should be 
> updated upon completion of DAD procedures if the created 
> state could cause inconsistent neighbor discovery behaviour."

sounds good.

> 
> Also in RQ4:
> 
> "For DAD optimization, a router MUST provide host requirements."
> 
> I think that what you're trying to indicate here is that
> 
> "A router which has been modified for DAD optimization
> should not interfere with hosts operating correctly
> in conformance with the host requirements document[HREQ]."

Originally, I considered router modification so as to support host
modification 
in stateful DAD Optimization. For example, router must provide DAD
optimization if some indicator as such flag, field are used for DAD
Optimization
on host.
 
" A router which has been modified for DAD optimization should
not interfere with hosts operating correctly. In particular, a router
has to meet the host modification for DAD optimization as well."

any comments ?


> in RQ7:
> 
> "At this case, all nodes MUST detect address duplication either
>       original DAD or DAD optimization.
> "
> 
> I think we have to be careful when we say all nodes must
> detect the duplicate address.
> 
> Since we've said that the systems must allow recovery from
> a duplicate event, this should take care of the other peer 
> nodes' neighbor cache state.
> 
> I think the main issue is to ensure that both the configuring 
> and the owner node (or both configuring nodes) are aware if 
> an address configuration attempt is made.
> 
> How about this text?:
> 
> "In the case that collision occurs, the configuring node MUST
>   inform the owner (or simultaneously configuring node) that the
>   configuration attempt has occurred.  In RFC-2462, this occurs
>   through the sending of the original Neighbor Solicitation packet.
>   Additionally, if intermediate neighbor cache state on peers
>   has been created, it must be repaired in conformance to RQ2."
> 

sounds good. I will make a change

> 
> I think things are shaping up quite nicely for this document.

I think so.



Regards

Daniel (Soohong Daniel Park)
Mobile Platform Laboratory, SAMSUNG Electronics