In an IPv6 future, how will you solve IPv4 connectivity?

Cameron Byrne cb.list6 at gmail.com
Mon Oct 11 05:35:54 CEST 2010


On Sun, Oct 10, 2010 at 5:48 PM, Truman Boyes <truman at suspicious.org> wrote:
> On 10 Oct 2010, at 11:43 PM, marcelo bagnulo braun wrote:
>
> El 10/10/10 16:10, Truman Boyes escribió:
>
> On 10 Oct 2010, at 10:05 PM, Roger Wiklund wrote:
>
> Let's say for arguments sake that the prophecy is true, and in late
>
> 2011/2012 a new user can only get an IPv6 address.
>
> Have you guys concidered/tested how you will solve these users
>
> connectivity to the IPv4 Internet?
>
> I guess NAT-PT is out of the picture,
>
> NAT64?
>
> DS-Lite?
>
> Also, as these new users are IPv6 only, how can IPv4 hosts communicate
>
> with them? 4to6 NAT?
>
> Thanks for your comments,
>
> /Roger
>
> I think the answer to this question depends upon the type of network (ie.
> mobile, internet application hosting, fixed line broadband, etc). DS-LITE
> would scale well, but would require CPE that obviously supported this
> feature.
>
> mmm, dslite, is about v4 hosts accessing to v4 servers (and using v6 in the
> ISP), so no translation is involved, so it would allow a v6 node to access a
> v4 server.
>
> Yes, however devices would need to be able to tunnel IPv4 inside IPv6. It's
> not translation, but the tunneling support is the key to traversing an IPv6
> network.
>
> NAT64 is simple, but it presents issues with tethering v4 devices among
> other issues.
>
> nat64 is the right tool for this particular problem, afaict
>
> I am not so sure about this being the right tool. Let's say that you have a
> mobile subscriber handset that is IPv6-only. The SP can turn on NAT64 and
> for the most part, the device would not need to care about IPv4. However, if
> you wanted to use the phone as a tethered modem/gateway on an IPv4 host, it
> would not work.
>

Service providers cannot support every OS forever.

I have tethered a IPv6-only + NAT64  Nokia 5230 to a Window Vista
laptop and it works fine for 99% of things i do. Your mileage may
vary, but it can certainly fit within some definition of a marketable
product to grandma, assuming she is not one of the grandmas that uses
skype.

Regional RFC1918 deployments + NAT44 can carry the IPv4-only users for
as long as needed, the size of the regions will keep going down and
the number of regions will keep going up.  It's not pretty, it's not
strategic, but will be required for some users for 10+ years.


IMHO, IPv6-only is inevitable.  There is no plan B.  Service providers
need to define products that work with IPv6-only today, build upon
what works today, and keep growing the services that work with
IPv6-only.  If you have not setup an IPv6-only lab, i encourage you to
do so.  I have documented some of my findings here:

http://www.youtube.com/watch?v=mdGv1WAYOEs
http://www.youtube.com/watch?v=czDvrNKKzpU

Of course, there are corner cases that do not work today, but WE need
to isolate them and either fix them or make them go away.  I see users
(eyeballs) being 1 or 2 types:  Those that require IPv4 and those that
work fine on IPv6 + NAT64.  Over time, the users that require IPv4
will get smaller.  For content, there is only 1 type: those that are
dual stack.

Regards,
Cameron


More information about the ipv6-ops mailing list