Comcast Activates First Users With IPv6 Native Dual Stack Over DOCSIS

Bernd Walter ticso at
Wed Feb 2 16:43:56 CET 2011

On Wed, Feb 02, 2011 at 08:44:12AM -0500, Jared Mauch wrote:
> On Feb 2, 2011, at 8:23 AM, Bernd Walter wrote:
> > It is counterproductive to speak about 2^64 addresses delivered with a
> > single /64 when we all know that all of the device classes with such a
> > density require LAN isolations, which is simply not possible with such
> > an allocation.
> You mean some people envision things being isolated in /64 lans.  I understand why people have that vision, this does not mean I agree with their view.

People always done this to isolate things.
There are other options of course, as proxy or bridging.
Proxy might be reasonable - especially light switches don't really
need to directly communicate with the public internet, but bridging
is a dirty thing.
The isolation has more requirements than just firewalling.
For example some automation networks e.g. for heating or light might
run on unreliable small bandwidth media, such as 802.15.4.
It is not a good idea to bridge it on the same network with your TV or
your filer.
The internet of things will come if you like it or not and it will
also come for home users.

> But for the purposes of a "first test user" with DS, honestly I don't see how your architecture arguments are relevant to the operation of some 3rd party network.

This is relevant if you sell home automation equipment or deliver
consulting for home users.
I personally don't have the problem in my own network, since I can
easily select an ISP supporting my personal requirements.
But I'm faced with customers who just have standard access only and
just what this standard will be become more and more dissapointing each
In my case I develop home automation devices and now I might have to
deal with regular renumbering and they might need to sit in the same
network as every other devices, but this is impossible.
IP renumbering for large lossy wireless networks can take many second
or even minutes.
You can't just share the same network with high bandwidth devices.
Automatic renumebring is Ok, but each renumber has a time gap until
every device has changed and do you want your light switch failing to
communicate with your LEDs once each day?

> Bring IPv6 to the consumer market is a huge task.  People have lived for years with NAT and other mind-bending technologies.  IPv6 represents the ability to get rid of some unnecessary NAT, I certainly agree.  I will be glad when I don't have to worry about these items myself and things can "just work".
> But just because you disagree with John and his first DS user setup does not preclude him or any other network from doing things differently.  I find the feedback useful to review, as with many conversations on all the lists that I consume.
> If you think what John is doing is a net-negative, I suggest providing that feedback to him directly.  I've found him very approachable myself when asking about what his plans are.  I don't like how long it takes, but then again, I understand large companies.

I don't disagree with him - he did it much better than many others in
that he doesn't use dynamic addresses and a single /64 will be enough
for most users.
I also assume that he is capable to supply larger ranges on request, so
as a default a /64 can be Ok - nevertheless he may want reconsider the
decision for a single /64 as it is no reason to be that sparse with
allocations - unlike with IPv4 noone kills you for assigning a /60 to
someone who is just using a single /64.

It is just that providers in general should also consider that the
expectations for IPv6 are different and that address space related
workarounds for IPv4 are not considered accepted for IPv6 as well.
Also fast deployoment technologies are welcome, but I'm feared that once
this is established ISP sleep for further 10 years again and we continue
to live with dirty hacks.
With a single /64 or maybe even dynamic one you have new restrictions
unless you switch back to the old ones.

I also do understand large companies, but we are talking about 10 years
with all major OS have IPv6 support.
Router support had been developing slower, especially hardware based
routing, but traffic had been low at this time as well.
Also this is not against John in person as many new technolgies such as
cheap DSL routers without IPv6 capabilities had been establied during
the 10 years, although allmost of the base on OSes which could do IPv6.
Many many more applications had been written without though about IPv6.
Until about 2 years ago all the problems have grown more and more.

B.Walter <bernd at>
Modbus/TCP Ethernet I/O Baugruppen, ARM basierte FreeBSD Rechner uvm.

More information about the ipv6-ops mailing list