Slashdot: UK ISP PlusNet Testing Carrier-Grade NAT Instead of IPv6

William Herrin bill at herrin.us
Wed Jan 16 22:22:55 UTC 2013


On Wed, Jan 16, 2013 at 2:53 PM, fredrik danerklint
<fredan-nanog at fredan.se> wrote:
>> ARIN and IETF cooperated last year to allocate 100.64.0.0/10 for CGN
>> use. See RFC 6598. This makes it possible to implement a CGN while
>> conflicting with neither the user's RFC1918 activity nor the general
>> Internet's use of assigned addresses. Hijacking a /8 somewhere instead
>> is probably not a great move.
>
> If I have calculated the netmasks right that would mean to set aside:
>
> 2001:0DB8:6440::/42
>
> for the use of 6rd service:
>
> 2001:0DB8:6440:0000::/64 = 100.64.0.0
> ....
> 2001:0DB8:647F:FFFF::/64 = 100.127.255.255

Sander already touched on this, but when implementing 6rd you'll want
*at least* 4 bits on the subnetting side of the IPv6 block associated
with each IPv4 address and you'll want that netmask to be evenly
divisible by 4. A /60 or a /56, not a /64.

In IPv4 your customer has a "DSL router," potentially with distinct
wired and wireless LANs running different RFC1918 address blocks. In
IPv6 each of those LANs will consume a /64, so he'll need more than
one.

Selecting a netmask evenly divisible by 4 has two major benefits.
First, it exactly matches one character in the written address. The
customer doesn't have ...:ABC4:* through ...:ABC7:*, he has
...:ABC*::. Second, each delegable RDNS zone takes up the same 4 bits
so the assignment will be right on an RNDS zone boundary.


>>> Even tough you have very good arguments, my suggestion would be to have a
>>> class A network (I got that right, right?) for all the users and only having
>>> 6rd as service on that network.

I assume you meant this a little differently than what you wrote here.
It wouldn't make any kind of sense to stand up a private IPv4 network
with no IPv4 Internet connection in order to facilitate IPv6 via a 6rd
deployment. For one thing it'd be a Rube Goldberg machine. For
another, I suspect you'd find it very challenging to acquire a
threshold number of paying customers for an IPv6-only network at the
moment.

Regards,
Bill Herrin


-- 
William D. Herrin ................ herrin at dirtside.com  bill at herrin.us
3005 Crane Dr. ...................... Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004



More information about the NANOG mailing list