Re: Re: [nemo] multiple CoAs registration

Hello Nicolas

> Hi Ryuji,
> 
> I carry on my remarks... hope it helps !
> 
> Ryuji Wakikawa wrote:
> 
> >Hello Nicolas
> >
> >thanks for comments.
> >
> >  
> >
> >>Hi Ryuji,
> >>
> >>I read your draft on multiple CoAs registration and I also think that we 
> >>really need to define a solution to use MIPv6 with a multiple interfaces 
> >>MN.
> >>
> >>Here are some comments on your draft:
> >>
> >>   - Why do you introduce the interface identification (IFID) ? It could 
> >>be sufficient to only register a CoA with the associated priority. 
> >>Therefore, when the MN has to choose the destination address for a HoA, 
> >>it can take the most preferred CoA, which is bound to the most preferred 
> >>interface of the MN.
> >>    
> >>
> >
> >Priority can be changed by some reason during communication, but IFID
> >can not be changed until MN want to change it.
> >
> Ok, I understand that. But what I mean is that I don't see the 
> requirement to have the IFID... In the Binding Cache, an entry could be 
> identified by both the Home address and the CoA. It is sufficient to 
> distinguish tow different entries. The priority field is then used to 
> choose the CoA, i.e. the interface. No need to have the IFID I guess.

I disagree.
ID is needed when CoA is changed.
If CoA1 is changed to CoA2, MN sends a BU with CoA2 to CN.
Before receiving the BU, CN has the binding entry which associate with
CoA1 and HoA, but BU carries new CoA2 in it. CN can not know which
binding cache entry should be updated for the BU. 

CN might create a new entry for the BU, because the pair of CoA1 and HoA
is different from CoA2 and HoA.

ID is enable CN to update the correspondent binding entry even if CoA is
changed.

> >
> >  
> >
> >>Anyway you use a one-to-one mapping between a CoA and a IFID, therefore 
> >>you don't need both, or I missed a point ?
> >>
> >>IMO, this is easier than all the mechanism to manage the IFID 
> >>(generation, maintenance...)
> >>    
> >>
> >
> >could be,  but I don't know why the maintenance and generation are so
> >difficult.
> >
> I don't say they are so much difficult, but they are not useful, since 
> the IFID is not necessary (from my point of view). Maybe you can explain 
> me when the IFID is useful ?

see above.
 
> >
> >  
> >
> >>   - If I well understand your draft, when the MN registers a CoA with a 
> >>CN, the MN has to first register the primary CoA. Then it can send a 
> >>Binding Update (after the RR procedure) to regsiter other CoA(s), with 
> >>the associated priority. However, you say that the lowest priority is 
> >>the preferred CoA/IF, and that the primary CoA has the priority 0. 
> >>Therefore, each CN will use the primary CoA with the MN. Then, you are 
> >>not able to perform load balancing, aren't you ?
> >>    
> >>
> >
> >Well, the draft does not mention about how to utilize multiple care-of
> >addresses. I don't think MIP is only way to manage policy sets to divide
> >flows to each CoAs. If you really want load-balancing, you need some
> >mechanism to maintain policy sets between MN and CNs.
> >
> >Load-balancing, for example, can be supported by upper
> >layers.
> >The point is that CN must be able to accept packets sent from multiple
> >.CoAs of MN. 
> >
> Ok I agree with you. But is it necessary to impose that a MN MUST 
> register the pCoA with the highest priority with CN ? I thnik it is 
> important that the MN does it with the HA, but I don't think it is 
> necessary with CN.

I got comments from WGchair whether I need priority field or not.
Policy can be treated as one of policy. I might remove the priority
field and keep the field as reserved.  

If there are no policies and priority values on CN, CN uses the first
matched BC entry for outgoing packets.

> If you have the choice to register the pCoA with a different priority 
> with different CNs, you let an open door for upper layers to decide some 
> load balancing policies.

This management is difficult. 
How does MN give different priority for each CNs. 
One of my intention is to eliminate these MN's decision from the draft.
I am interested in the policy management and notification, but it should
be described in a separate draft.

anyway, the issue of priority field is in my todo lists. 
I appreciate if anybody has comments on this

regards,
ryuji

> >
> >  
> >
> >>   - Last point: If you return home with your primary interface (you 
> >>connect to your home network with P-IF), do you have to de-register all 
> >>CoA (P-CoA and NP-CoA) with both your HA and CNs (see section 7.8) ?
> >>    
> >>
> >
> >You are right. I personally got the same comment. 
> >I just keep protocol simple. 
> >I will update the draft before the next meeting.
> >
> >thanks again!
> >
> >regards,
> >ryuji
> >
> >  
> >
>