On 12/3/09 1:54 AM, "bo zhou" <[email protected]> wrote:
> 
> [Bo] I agree "DS-lite makes E2E transparency argument." But left part I do not
> agree. Host-based translation can make the applications running through the
> transport network transparency, never care what kind of transport network
> (IPv4, IPv6) running. Considering of E2E concept, sort out issues at the end
> is better than in the middle of network.
> By the way, I think there is a tricky point in DS-lite, DS-lite can assign
> unlimited IPv4 address to the host, the concept behind this point is very
> similar as A+P: save IPv4, I believe DS-lite is a good solution to save IPv4,
> but I am not sure it is a good solution for IPv6 migration and deployment.
> People use DS-lite solution never want to update their service to support
> IPv6, because no risk of IPv4 address exhaustion exist if DS-lite running
> well. 
> 
> ===> Bo: 
> From a service provider perspective, what matter is NOT which flavor of IP I
> like or don¹t. What matters is the service
> I can offer to my customers. It is true that today, a v4 service has much more
> value than a v6 service. Things will certainly change, but right now, this is
> the case.
> 
> Now, I want to refute your argument than running DS-lite, you¹ll never want to
> update the service to support IPv6.
> First and foremost, DS-lite runs over IPv6. That is you MUST deploy IPv6 in
> the access network to deploy DS-lite.
> Second, this is the opportunity to break the chicken and egg problem. Now, you
> have a real v6 access network in place, you can have a serious conversation
> with application vendors & content providers  to port their stuff to IPv6.
> Third, every packet that moves over the now enabled IPv6 network is a packet
> that does not move over the DS-lite NAT AFTR IPv4 box. Ie, the more IPv6
> traffic we get, the less reliance we have on that NAT infrastructure and we
> can expect to cap the cost.
> 
> Is it a slower route than Œmandating¹ IPv6 everywhere right now? Sure. But at
> least it is a doable thing, that makes business sense.
> In the last 15 years, the argument ³let¹s just do IPv6² has not been very
> successful...
> 
>    - Alain.

_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires

Reply via email to