Re: New Version Notification for draft-koodli-ipv6-in-mobile-networks-02

[Rajeev Koodli <rkoodli@cisco.com>]

On 4/19/10 10:02 AM, "Rémi Després" <remi.despres@free.fr> wrote:

> 
> Le 16 avr. 2010 à 18:45, Koodli, Rajeev a écrit :
> 
>> The ID states that the IETF dual-stack model is readily applicable in mobile
>> networks, and that's the likely deployment model for a majority of mobile
>> providers today.
> 
> "Likely", yes, but not "recommended" as IPv6-only in the text as is.

I will make it explicit that dual-stack is the IETF recommended model.
However, I do think that IPv6-only deployments in mobile networks need to be
*looked into* and *encouraged*. See below..

>> Separately, IPv6-only deployments may be relatively few, but need to be
>> looked into.
> 
> "Looked into", I do agree, but not prematurely "recommended".
> 
> Among problems I see, one is IMHO very important:
> - E2E transparency is a key feature that IPv6 restores.
> - If DNS64-NAT64 synthesizes IPv6 addresses that hosts don't distinguish today
> from real IPv6 addresses, this destroys the IPv6 guarantee of e2e
> transparency.
> 
> In my understanding, until all concerned applications do recognize
> IPv4-embeded IPv6 addresses from real ones, DNS64-NAT64 will be harmful to
> IPv6, and should be advised against.
> 
> Dual-stack deployments (with NAT44 CGNs and guaranteed IPv6 e2e transparency),
> offer on the other hand a clean IPv6 service that IPv6-only deployments with
> DNS64-NAT64 can't offer.

We are trying to get to IPv6 which offers many things (including e2e
transparency), but we have to face the fact that it is not a clean switch.
Some folks argue that investments in NAT44 would "silo" IPv6 into some
select usage only. Other folks do not wish to have to worry about private
IPv4 addressing at all; this can be argued as an impending acute problem in
mobile networks. Yet another consideration is the need to maintain multiple
PDN/PDP if you do separate IPv6 and IPv4. IPv6-only deployments offer *a*
choice here. Whether it comes with a clean switch is moot. So, there are
tradeoffs.
 
Having said this, dual-stack is the easier choice for many. The draft BTW is
not preferring one over the other.


> In sec 3.3, you have (asterisks added):
>    " Hence a service provider may
>    choose to enforce IPv6-only PDN and address assignment for their own
>    subscribers in their Home Networks, see Figure 1.  This is feasible
>    for the newer MNs when the provider's network is "IPv6-ready", which
>    means the network is able to provide IPv6-only PDN support *and
>   IPv6 - IPv4 interworking*  for Internet access."
> 
> I therefore understood the considered "IPv6-only" networks as having
> DNS64-NAT64s (not as having no IPv4 connectivity at all).
> 

It's both: no IPv4 connectivity, and IPv6 - IPv4 interworking.

Dual-stack deployments do not have either of these.



>> ...
>> I could reword the last sentence to: "For instance, if an LTE deployment is
>> IPv6-only, a roamed 3G UMTS network may not offer IPv6 connectivity"
> 
> Isn't it true that the lack of IPv6 connectivity of some 3G UMTS networks is
> independent from whether LTE networks are dual-standard or not?
> 

Yes. However, the point here is that if an LTE deployment chooses to be
IPv6-only, a roamed 3G network may not offer IPv6 connectivity, which means
the home network (that offers IPv6-only service) needs to be able to provide
IPv4 connectivity to such roamed users.


>> 
>> Dual-stack is the recommended model.
> 
> Making it very clear will then be an improvement (IMHO).
> 

Check.



>> "In summary, IPv6-only deployments should be encouraged while considering the
>> roaming and applications issues"
> 
> I still can't agree on this because:
> - The market of real IPv6-ony is too limited.

Well, someone has to start. There are arguments on why some want to do
IPv6-only in mobile networks.


> - IPv6-only + DNS64-NAT64 is harmful to IPv6 e2e transparency.

See comments above.

> - With simple additions in LTE MNs port-restricted public IPv4 addresses can
> be made available across IPv6-only access networks, which reconciles IPv6 e2e
> transparency and IPv4 connectivity with an IPv6-only routing plan in LTE
> access networks. (About this, see also what I answered to Cameron Byrne in
> http://ops.ietf.org/lists/v6ops/v6ops.2010/msg00609.html).

I understand. Making changes to MNs is another story. I reckon 3GPP would be
a good place to bring this kind of proposal.

> 
> IMHO, if your draft can be a great informational document if relieved of
> general "encouragements" to deploy IPv6-only (at least as long as it
> implicitly goes with DNS64-NAT64), and if noting the conflict between IPv6 e2e
> transparency and DNS64-NAT64.

Thanks. 
However, I do believe that there is a case for IPv6-only deployments from a
variety of angles, including - private IPv4 address exhaustion, side effects
of overlapped addresses etc., investments in NAT44, speedier transition to
IPv6 and so on. This does have costs associated with it, as you point out
above.

It may be worthwhile to refer to NAT64+DNS64 considerations to an
appropriate BEHAVE document?

Again, I would point out that the ID is not intending to prefer IPv6-only
over Dual-stack. The latter is the default, the former should be a choice,
at least for some. I will clarify..

Thanks,

-Rajeev
  


> 
> Best Regards,
> RD
>