Presuming IPv4 (was Re: Routing of 2002::/16)

[Nick Sayer]

Michel Py wrote:

 >>> From: Chuck Yerkes [mailto:chuck+6bone@snew.com] This may be a
 >>> reasonable assumption at the moment, but in the context of an
 >>> experimental and experimenting network (the 6bone) and, more,
 >>> in the interests of preparing for a far more 6-aware world, I'd
 >>>  like to presume instead that there are many 6-only networks
 >>> with no 4 network.
 >>>
 >
 > Dave's original text had the word "customer". Today, and for a
 > while, the best advice you can give to a customer is that they need
 >  both v4 and v6. V6 only networks  are good for experiments, but I
 > would not run a business on them. Would you be able to read this
 > email today if you were v6 only? Would you risk loosing customers
 > that have transition mechanism problems for the pleasure of being
 > v6 only?

I don't know if I would go as far as that.

I believe that an enterprise needs a handfull of IPv4 addresses for 
things like web services, mail relays and NAT/NATPT endpoints. But if 
you're going to deal with the Internet via NAT, doing so via NATPT isn't 
any different, except for the fact that you have an opportunity to 
communicate with v6 sites without translation.

 > Let's be serious. The 6bone is a great place to test, but the day
 > when Microsoft, Cisco, eBay, Amazon, Etrade, Travelocity and Cnn
 > are moving their web sites to v6 only is years away.

Perhaps, but the day that they add AAAA or A6 records to their DNS is 
perhaps a bit closer. In the meantime, I can reach them with NATPT 
without having anything except the NATPT gateway configured with IPv4.

 > And I don't
 > envision an Internet where you can pretend that Windows does not
 > exist any time soon either.

There exist enterprises without Windows anywhere to be found. If you 
provide services for external sites, they will clearly need to have 
routable IPv4 addresses for now. But that's not the whole story here.

 > In the meantime, the reason I was mentionning that it would be
 > preferable to have the customers have their own 6to4 solution is
 > this:
 >
 > In the next three years, you will update the software that runs the
 > 6to4 relay many times. People that have been doing this for a while
 >  will agree that there is a good chance that one of these many
 > upgrades will introduce a new bug that will disable the 6to4 relay.
 >
 > If your design is that your customers rely on this 6to4 relay, when
 > you break it, all the customers are screaming at the same time,
 > which experienced network administrators try to avoid.
 >
 > On the other end, if customers have their own 6to4 relays on their
 > own routers, not only they will not break all at the same time, but
 >  when they do break, it's not your fault.

Replace '6to4 relays' with 'e-mail' and you can make the same argument. 
They are adding value to their service, which I applaud. It makes their 
service more valuable. As for the idea that they're going to flash some 
software into a router and suddenly hose up their service, the same 
argument applies to any other aspect of their business. Which is why you 
  either thoroughly test such things before hand or have a contingency 
plan to quickly migrate the service elsewhere or back the upgrade out 
should the need arise.