RE: Open issues list? [Re: New (-02) version of IPv6 CPE Router draft is available for review]

["Stark, Barbara" <bs7652@att.com>]

>It's the observable behaviour that matters.
>
> Antonio Querubin

I really agree with this statement, and wonder if it would be more
useful to describe externally observable behavior, than to try to
specify implementation (loopback interfaces and such). And identify what
external stimuli the CPE router needs to be able to respond to, and what
those responses are. I look at CPE routers as black boxes. Either they
behave the way I want, or they don't. What they do on the inside to
behave that way, doesn't really matter to me. I can't test that -- I can
only test what I can observe.

That is, when a CPE router's WAN interface is connected to the access
network, the access network sees messages coming from the CPE router,
and sends messages in return (router advertisements and such).

Given that we appear to agree that the access network dictates whether
stateful or stateless address assignment happens, and whether or not it
hands out a global address (i.e., whether or not the access network uses
the numbered or unnumbered model), the CPE router that expects to work
on all types of access networks needs to be prepared to respond
appropriately to all of these cases. A CPE router that is specifically
designed to only operate on a particular network doesn't need to be
prepared for all of these cases, but I'm not really interested in
specifying that CPE router in the IETF. As I've stated before, my
biggest concern are the retail CPE routers with Ethernet WAN ports.

The non-network-specific CPE router must be prepared to act
appropriately when it gets a GUA (through SLAAC or stateful DHCPv6). It
must also be prepared to act appropriately (from the perspective of an
external observer) when it does not. In my opinion, a
non-network-specific CPE router must also boot with LAN coming up
without dependency on the WAN (the LAN before WAN boot mode). 

If the "loopback interface" somehow appears different to the access
network, and the access network needs to know this, then this interface
is relevant. If the access network (or the LAN, either) cannot tell that
there is a loopback interface, or doesn't care, then such an
implementation isn't relevant.

What I do care about is the GUA that the CPE router uses when it
initiates IP traffic to other global addresses (not within its assigned
prefix). I think it should use an address from the prefix. If the access
network didn't give it a GUA (unnumbered model), then I think it *must*
use an address from the prefix. I don't really care how this is
implemented internally, so long as there is agreement that it is
possible and acceptable that this be able to occur. I don't care if this
address is somehow connected to the "WAN interface", or not. So long as
the router uses it.

Barbara