[arin-discuss] ipv6 technology supplier phone bank?

[michael.dillon at bt.com]

> > Crazy question. Anyone deploying IPv6 to the home user is going to 
> > make sure that those IPv6 customers will get access to all of the 
> > Internet. 

> > Flashable home gateways, especially those based on Linux, 
> > can be easily upgraded to support IPv6, so not all home 
> gateways will 
> > need to be replaced.
> 
> How does that get you to the 50% of the Internet that isn't on IPv6?

Two separate answers to two questions. Obviously, the gateways are not
necessarily the way to get access to the IPv4 Internet. Yes, that means
that ISPs will have to invest in 6to4, NATPT, tunnel servers and 
whatever else comes along.

> Are CPE vendors going to provide code for those gateways, or 
> are ISPs going to write code and take support calls?

I expect that the Open Source community will write the code, 
the CPE vendors will adopt the code and support it, and the
ISPs will choose whether they want to take the support calls 
themselves or foist that onto the vendor.

> I strongly recommend that every ISP look at its customers' 
> favorite sites and find out the IPv6 plans for each of them.  

That should be part of the IPv6 sprint. It is not just the vendors
that we use in our own businesses that matter. In fact, given how
far along some key vendors are like Cisco, Juniper, SUN, etc.,
you could argue that it is more important to reach out to the 
suppliers that our customers use, including the websites that they use.

> > They can still switch providers and receive a comparable 
> IPv6 service. 
> > Competition will not disappear.
> 
> You think IPv6 service will be comparable to Internet access 
> with unCGNed IPv4?

People don't judge ISPs based on some study of percentage reachability.
They judge them based on the quality of the experience accessing a 
handful of their favorite sites. So yes, IPv6 service will be comparable
to unCGNed IPv4. Both will be unable to reach certain sites when
new IPv6 services come online.

Why would a site launch with only IPv6 access? Perhaps because they
want to deal with steady controlled growth without risking a huge
onslaught
of new users in the space of a week. Perhaps because they want to try
some
new features in a controlled trial on an IPv6 version of their main site
and don't want to spend too much on supporting the non early adopters.

> I complete agree, but I think Tony's looking for more detail.
> Do home gateways need to support both SLAAC and DHCPv6, as 
> both a client and as a server?  Do they need to provide 
> stateful DHCPv6?  Do they need to be DNS servers, or relay 
> dDNS updates?  If somebody daisy-chains a wi-fi router behind 
> an edge gateway, how does it get a prefix delegation from the 
> gateway's prefix delegation?  
> 
> In other words, people need to be deploying IPv6 now, so they 
> can find out what they don't know, and have time for vendors 
> to fix bugs in features they need.  Simply ordering a heap of 
> IPv6 and expecting to be ready isn't enough.

I wholeheartedly agree. And if you think about the timing of
all these stages of testing, both in the vendors and the ISPs,
and add them all up, they are getting pretty darn close to the
projected dates for IPv4 runout. It's not too late yet, if people
get their butts in gear now. But in another 6 months or a year
it will be too late. Then the only way to get IPv6 out in time
will be to take shortcuts and pray that you don't make a mistake.

--Michael Dillon