Re: ND-proxy applicability and loop-prevention

[Ralph Droms <rdroms at cisco.com>]

At 12:09 PM 3/29/2004 -0800, Erik Nordmark wrote:
> > The problem is that with the same trouble it takes to fully delegate a
> > /64, the ISP could do a /48 as well.  That is a good thing also, of
> > course.  My worry is that the ISPs end up doing nothing unless they
> > have simple enough means available.
I know that there are service providers in Japan that plan to
use DHCP PD to provide /48 prefixes to customers.  Therefore,
DHCP PD is apparently "simple enough" that some ISPs do not
consider it a barrier to deployment.

Do we have any examples of ISPs that have reviewed DHCP PD
and rejected it as not simple enough?

> My worry is that ISPs would end up doing nothing because of the existance
> of nd proxy.
> Hence I conclude we should make prefix delegation work and make the
> implementations of it as easy as possible to use.
> Then the ISP can decide whether to delegate nothing (and customer gets
> a single address) or delegate /64, /48.
I agree that we should continue to work on improving prefix delegation
(if, indeed., it needs improvement).

> > I mean, the message could also be "just give every customer a
> > delegated /48.  If that is not possible, advertising a /64 on the link
> > is better than nothing."
>
> Yes, but that requires a non-robust against loops nd-proxy complexity
> to be added to the architecture.
> In the big picture this is a distraction and not a help.
>
>   Erik
I agree here, as well ... we have a solution that works, has been
accepted as a standard and is being deployed.

- Ralph



--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------