On 9/10/2010 5:59 AM, Jari Arkko wrote:
Some of the discussion has gone into the history of IPv6 design, what
configuration model was intended by the original designers as the
right one, and so on. I would suggest that while that's interesting,
it may be secondary to what we are discussing here.
But it's not. My main operational experience, as well as my main IETF
experience as a participant has been in the DNS. We DNS folk have a long
history of "Let's find a way to make new stuff work with legacy systems"
that has led us to the giant mess we're in now. (Unfortunately) it's
still early days in the history of deployment of IPv6. We _really_ want
to get this right NOW. Adding more kludges so that we can "Just get it
deployed" is actually going to make life (and future deployment) harder
down the road, not easier.
Suresh wants to support a particular type of a deployment, and it
doesn't help him if some design is the "right" one or matches with
our *desires* of what hosts and other devices should be doing or
implementing. The primary concern should be that we have something
that works.
Wrong wrong wrong wrong wrong. Sorry, but maybe if I say it enough times
in this post I'll stop having to say it over and over again in
subsequent posts. I know I'm talking out of both sides of my face here,
since I'm usually the one who is complaining about the IETF not
listening to the operators (especially in IPv6) but this is one case
where the operator is just plain wrong, and what they want to add to the
specification is a not only a bad idea, it's a dangerous one.
In particular, we need to worry about what random hosts
do in these networks, as we cannot dictate what devices will exist in
people's homes. And we cannot upgrade or change existing
implementations. This applies both to the problem that started this
discussion, i.e., hosts giving up on sending RSes but also to their
support of DHCPv6. As an example, while DHCPv6 is available on Linux,
none of the IPv6 Linux devices that in my network have come with
DHCPv6 on by default. (You can of course ask me to change my configs,
but asking my mother to do the same may not be feasible.)
Your mother runs linux? Cool!
Host support is important because that is an area where neither the
IETF, any single vendor, or the DSL operators have any easy way to
change the situation. But it is of course by no means the only
constraint. The operators have their issues as well. Even though they
are in control of their own networks, they probably want to keep
their underlying L2 network architecture as it is, despite
introduction of IPv6 or other new features.
Two responses. If we can't expect the hosts to be changed in order for
this to work, how do we expect them to send clever new RS messages even
if the draft is adopted (or perhaps I'm misunderstanding something)? And
I don't know if it's common in other countries or not, but every
broadband account I've ever signed up for in the US has come with a
dis[kc] of special software that I "had to install" to make my system
able to access the intartubez. Now back in the 90's when these things
came on floppy disks there often was some sort of mandatory sign-in
software that I couldn't get on line without. Since I was dual-booting
FreeBSD at the time, and the software was only available for windows, I
had to figure out a solution for myself because my OS was not supported.
Nowadays these things come on CD, and I've long since learned NOT to
install them because they are overwhelmingly likely to do Bad
Things(TM). But it does make one think that there may be a solution to
the "Random host doesn't do DHCPv6, and/or doesn't do it by default"
problem; even if one doesn't like Woj's CPE-related solution (which I've
already said I do think is the right answer).
Doug
--
... and that's just a little bit of history repeating.
-- Propellerheads
Improve the effectiveness of your Internet presence with
a domain name makeover! http://SupersetSolutions.com/
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
