Re: [BEHAVE] ***SPAM*** 5.548 (5) Is nat46 worth researching？

[Iljitsch van Beijnum <[e-mail]>]

On 28 apr 2011, at 14:11, buptnoc wrote:

>     As described in draft-ietf-behave-v6v4-framework-10#section-2.4 ， we need nat46 translator. 
>     But, do we really need this scenario？Is it worth to deploy this scenario?

>     In fact, this scenario appears when we have v4-only client and v6-only servers

My opinion is: no, this is not worth the trouble. We know that NAT46 is a hard problem, and it's unlikely a solution would be very robust. Because of lack of IPv4 addresses, a relatively small pool of v4 addresses would have to map to all possible v6 addresses, which means that the mappings have to be highly dynamic. But addresses are cached in many places, including often for a long time in applications. Having different applications react differently to NAT46 would be a big deployment problem.

I would recommend (apart from upgrading to IPv6) deploying HTTP and HTTPS proxies, as those will allow HTTP and HTTPS from IPv4-only clients to IPv6-only servers (or the other way around!) and in principle, it's possible to modify any TCP-based application to work through an HTTPS proxy, as those are basically TCP relays.

It should be possible to make an automatic proxy configuration so that a browser only uses the proxy to reach IPv6 destinations and connects to IPv4 destinations directly. However, I haven't tried this myself yet.