Re: WPD-6, WAA-8, and WAA-9 of draft-ietf-v6ops-ipv6-cpe-router-04

[Philip Homburg <pch-v6ops@u-1.phicoh.com>]

In your letter dated Mon, 5 Apr 2010 21:44:17 +0200 you wrote:
> >>> So, what is wrong with that scenario?
> >>
> >> Section 12.1 of RFC 3633 forbids it:
> >>
> >> "the requesting router MUST NOT assign any delegated prefixes or subnets
> >> from the delegated prefix(es) to the link through which it received the
> >> DHCP message from the delegating router."
> >
> > I know, but why?
> >
> > I always see a MUST NOT as something that would cause major harm or
> > interoperability problems. But the DHCP PD RFC just says 'MUST NOT' without
> > giving any reason what so ever.
> >
> > So maybe it is time to revisit that requirement instead of using that to
> > force the weak host model onto CPEs.
> 
> the reason for that "MUST NOT" is to prohibit path of the delegated
> prefix from being used on the DR - RR link. if the DR chose to use
> part of the delegated prefix, that would be hard for the RR to
> detect and could potentially conflict with the RR's own use of the
> delegated prefix. "delegation" implies that the DR transfers control
> of the prefix to the RR.

I don't understand what you are saying. The quoted text says "the
requesting router MUST NOT assign [..]", so this a requirement on the 
requesting router.

I don't think the DR can do anything with part of the delegated prefix 
anyhow.

I'm not proposing that the DR assigns a subset out the delegated prefix
to the link, that would not work. I'm proposing that the requesting router
assigns just for itself an address from that prefix to it's WAN interface.

> this doesn't necessarily prohibit the RR from doing what you suggest.
> but, I haven't understood what problem you are trying to solve.
> could you expand on that?

The current text in draft-ietf-v6ops-ipv6-cpe-router-04 mandates the
weak host model: if the CPE gets a delegated prefix with an unnumbered WAN
link then RFC-3633 prevents it from assigning an address out of that prefix to
it's WAN interface.

I prefer to avoid unnumbered links. I think it is much less confusing if 
source addresses of error ICMPs are always addresses from the interface that
generated the error.

So I'd like the CPE to be able to assign an address from it's prefix to it's
WAN interface.