Re: [lisp] Draft-kouvelas-lisp-map-server-reliable-transport WG Adoption

[Dino Farinacci <[e-mail]>]

Would both be active at the same time? 

If so, control-plane would be much better since one or the other RLOCs would be used. So old cachers use one address and new cachers use the other address. Would work well. 

With predictive-RLOCs, dup packets would be delivered to application if both addresses active at the same time. 

Dino

> On Apr 23, 2022, at 2:21 PM, Robert Raszuk <[e-mail]> wrote:
> 
> 
> Hi Dino,
> 
> I am referring to SSC Mode 3 which LISP application can choose to use and where with single radio you get two active IP addresses on a device allowing you time to register with LISP mapping plane new RLOC
> 
> https://www.techplayon.com/ssc-modes-session-and-service-continuity-in-5g/
> 
> 5G and wifi would be exactly the same. 
> 
> Mobility while on the plane or train is usually behind NAT anyway so you need to keep data plane active,  
> 
> So practically what's left ? 
> 
> Best,
> R.
> 
> 
>> On Sat, Apr 23, 2022 at 11:03 PM Dino Farinacci <[e-mail]> wrote:
>> > I am actually referring to the case where the control plane during the handoff registers new RLOC while the active/old one still works. Nothing to do with the data plane. 
>> 
>> You can register the new RLOC and new cachers will use it. Old cachers will use the old one. Now the question is, where is the EID, it can be the old, the new, and in transit (meaning offline).
>> 
>> If you are an EID that is at both RLOCs at the same time, and want to migrate from one (removing one of them), then this works per specs.
>> 
>> > That is what I meant by "radio overlap". 
>> 
>> There is no such thing as overlap unless the EID has two interfaces each with an RLOC. But that is not the interesting handoff case. I know cells overlap but most phones have ONE cellular interface and therefore one RLOC at a time. I wish this could change, but the mobile OSes aren't there.
>> 
>> > Then there will be time where more then one RLOC is active and the old one can get pruned after the new one is tested as working fine. 
>> 
>> You have to define the details of "active". That is what brings interest to this discussion.
>> 
>> > Would that not be a better overall option ? 
>> 
>> Its not that simple.
>> 
>> Dino
>> 
>>