Re: [Softwires] DS-Lite vs. 4rd

[Suresh Krishnan <[e-mail]>]

Hi Edward,

On 10/21/2015 08:29 AM, Edward Lopez wrote:
> I apologize if this has been thrashed out in the past.  In looking as implementing DS-Lite support, it appears that the need to include an additional tuple of information on the IPv6 B4 address of the CPE is cumbersome to NAT performance and tunnel capacitance, as many HW accelerated NAT engines exist without this extra tuple.  It would appear that by splitting the AFTR into two functions, 4in6 encapsulation & NAT(CGN), we can overcome scaling and performance issues of DS-Lite.
>
> However, the issue of overlapping endpoint subnets supported internally by the CPE leads to the issue potentially supporting NAT44 on the CPE, to support stateless encapsulation of returning IPv4 packets into IPv6 by the AFTR.  Section 4.2 of RFC-6333 states that CPE devices ‘should not’ perform NAT44, but that’s not the same as a ‘must not’
>
> But as you craft this solution out, you begin to realize that you are re-creating the majority of 4rd, RFC-7600.  However, 4rd is currently an experimental standard.
>
> My questions:
>
> -	Has anyone implemented or considered implementing DS-Lite with CPEs performing NAT44?

Have you looked at RFC7596 (Lightweight 4over6)? Its main goal was to move 
the NAT to the B4 (CPE) from the AFTR.

Thanks
Suresh