Re: [v6ops] I-D Action: draft-ietf-v6ops-6to4-to-historic-06.txt

[Jeroen Massar <[e-mail]>]

On 2014-10-30 21:44, Keith Moore wrote:
> On 10/30/2014 04:41 PM, Gert Doering wrote:
>>>> Before you get a 6to4 tunnel, stick to IPv4-only.  6to4 tunneled IPv6
>>>> > >to the "wild Internet" (unlike "to your friend who also uses
>>>> 6to4, no
>>>> > >gateway involved") will be SO much worse than "just use native IPv4"
>>>> > >that you're doing nobody a favour doing this.
>>> >It's a mistake to assume that the reason for using IPv6 is to talk to
>>> >the "wild Internet" (especially if you define "wild Internet" as the
>>> set
>>> >of hosts that talk  IPv4 and might also happen to talk IPv6).   The
>>> >motivation to use IPv6 is driven by different applications than IPv4
>>> was.
>> So in which way does this draft impact your ability to talk IPv6 between
>> your nodes?
> It impacts my ability to talk IPv6 with any other nodes for which 6to4
> currently works, which are not necessarily "my" nodes, nor nodes that
> also support IPv4.   To give one example, I've found that video
> conferencing over 6to4 can work in situations in which STUN-based NAT
> traversal fails.

You are saying you can transport proto-41 packets through a NAT setup,
that while STUN does not work?

I would really love to see the details. As NATs that know about proto-41
or are able to stick it to a specific endpoint are rare.

In most cases to get proto-41 working behind NAT you will have to force
the NAT box into "DMZ mode" (or whatever the flavor of the name is). And
even then, typically they do not do proto-41.

The DMZ trick will fail anyway, as you cannot teach your local host to
use the local IP address (eg 192.0.2.1) and configure it to listen to a
6to4 address of 2002:<public>:<ipaddr>::<something>

Thus I am *really* wondering how that is supposed to work.

Unless you mean you let the NAT box do the 6to4, as then, indeed things
get better. That requires also the other host to do 6to4. But in that
situation you will have an up-to-date enough host that will properly
handle STUN, uPNP or NAT-PMP and should thus not have any issues with NAT.

You will have a much better chance using Teredo in this situation (which
is what Xbox One does btw :)

Or much better: just use TSP or AYIYA, then you have determinism on your
side.

Greets,
 Jeroen