Re: [v6ops] The need for local-ipv4 socket transition solutions -- NAT64/DNS64 remains insufficient

[Mark Andrews <[e-mail]>]

In message <[e-mail]>, Xing Li writes:
> Ross Chandler 写道:
> >> On 5 Apr 2015, at 10:38, Xing Li <[e-mail]> wrote:
> >>
> >> Ca By 写道:
> >>     
> >>> Conclusion:  IPv4 sockets need to be supported on hosts that operate in I
> Pv6-only networks.
> >>>       
> >> Fully agree, based CERNET2's 10 years IPv6-only backbone experience. We ha
> ve the following observations.
> >>
> >> (1) If it is the IPv6-only network and IPv6-only applications (IPv6-only s
> ocket), then nobody will use it, except for the demonstration.
> >> (2) If it is the IPv6-only network with single IPv4/IPv6 translation and I
> Pv6-only applications (IPv6-only socket), then somebody will use it, but they
>  are not the majority (less than 5%).
> >> (3) If it is the IPv6-only network with double IPv4/IPv6 translation and I
> Pv4/IPv6 applications (IPv4/IPv6 sockets), then everybody will use it happily
> .
> >>
> >> So I think the transition path should be moving from double translation to
>  single translation and eventually to IPv6-only.
> >> Regards,
> >>
> >> xing
> >>     
> >
> > Once IPv6-only with double IPv4/IPv6 translation is available the next stag
> e to reach is IPv6-only. IPv4 literals will be with us for as long as IPv4 is
> .
> >   
> 
> If all the users on the Internet can access both the IPv4 (via double 
> and single translation) and IPv6 (native) contents, who cares if the 
> IPv4 literals will coexist with IPv4 and IPv6 for a long time?
> 
> > Similar to ISPs dropping services they don’t have a natural advantage in pro
> viding (Usenet, web hosting, email, in fact anything not tied to the access n
> etwork) they eventually won’t bother to operate NAT64.
> >   
> 
> If the ISP lacks the public IPv4 addresses, the justification is to compare
> 
> (1) (stateful or stateless NAT64 + [host based] double translation)
> (2) (NAT44 + dual-stack)
> (3) NAT44-only

(4) DS-Lite

Additionally ISP's don't need to operate the translation/encapsulation
services themselves.  464XLAT and DS-Lite both can be done elsewhere
on the net.  Eventually IPv4 will be a "service" you get from a
niche service provider not your regular ISP.

> Based on 10 years IPv6 operation experience of CERNET2, I think (1) 
> costs less than (2) and (1) has more features (IPv6) than (3). Therefore 
> (1) is the way to go.
> 
> Regards,
> 
> xing
> 
> 
> > Ross

-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: [phone]                 INTERNET: [e-mail]