On 21-mei-04, at 19:00, Pekka Savola wrote:
All the services IETF servers offer are purely client-server based.
There is no significant technical advantage that I could see in making
them IPv6-enabled, because all such services are very usable with
IPv4. On the other hand, doing so would just strengthen the illusion
that wide-scale migration of all IPv4 services to IPv4/IPv6 is an
important short-term goal.
While the importance of making any specific website or other service
IPv6-capable is debatable, I certainly believe migrating these to
dual-stack is an important goal. (I'm not saying short-term, as there
is no specific deadline, but this doesn't mean waiting forever is a
good idea.)
The reason for this is very simple: we sorely need operational
experience. The proof of the pudding is in the eating and all that. The
good part about making websites v6-capable is that people who use
IPv6-capable browsers on IPv6-capable systems with IPv6 connectivity
then get to see how well this works, without undue interference to IPv4
reachability of the same website.
The idea that it would be possible to start using IPv6 for services
that can't work well with IPv4 is a mistake, as breaking the
chicken/egg cycle is much, much harder in this case. If it is
completely impossible to deliver the service over IPv4 and living
without it isn't an option, then IPv6 adoption will happen. But guess
what: there is nothing you can do with IPv6 that you can't also do with
IPv4. It just takes more effort.
What we need to do is show that providing connectivity using IPv6 is
actually easier than doing the same using IPv4. This isn't entirely the
case just yet, but we're getting there. And an important aspect of
showing that something is easy is simply doing it.
Unfortunately, pretty much everyone seems to assume we'll be running
dual stack until the cows come home, and obviously running dual stack
is more work than just running IPv4. A bit more focus on the time when
we get to remove IPv4 from 90% of the network would be in order.
See Keith Moore's excellent write-up:
http://www.cs.utk.edu/~moore/opinions/ipv6/dubious-assumptions.html
Sorry, I'm not impressed. I'm not sure whose assumptions Keith tries to
debunk here, but just replacing them with his isn't much of an
improvement.
For completeness (not necessarily agreeing with everything, probably
want to skip the IPv6 introduction):
http://www.potaroo.net/papers/isoc/2003-01/Waiting.html
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf