RE: Comments on draft-wbeebee-ipv6-cpe-router-01.txt

["Hemant Singh (shemant)" <shemant@cisco.com>]

 -----Original Message-----
From: Antonio Querubin [mailto:tony@lava.net] 
Sent: Friday, July 18, 2008 11:57 AM
To: Hemant Singh (shemant)
Cc: Stark, Barbara; v6ops@ops.ietf.org
Subject: RE: Comments on draft-wbeebee-ipv6-cpe-router-01.txt

I wonder, for those who feel a loopback interface is important, is that
because you want the CPE router to be pingable regardless of any PHY
interface status?  

<hs> 
That is one reason senior (savvy) folks like Remi on this mailer gave.
The other reason is equally critical. Some folks requested they only
wanted to assign a link-local to their WAN interface and use a Loopback
interface to assign a global IPv6 address to. Since the device is a
router, spawning any virtual interface on the router is no big deal for
us. We said fine. Then as Ole pointed out today, the Loopback interface
is also a solution for folks like AT&T who want to assign a global
address to the "WAN" interface from the IA_PD that was doled out by the
SP. As Ole said, RFC3633 prevents a RR like the CPE Router to assign any
address from the IA_PD to an interface that the DHCPv6 REPLY ingressed
on. Any DHCPv6 REPLY has to get to the CPE Router via the physical WAN
interface. The Loopback interface gets around this restriction of
RFC3633. However, if AT&T doesn't like using a virtual interface like
the Loopback interface, then I said they have cannot assign an global
address to the WAN interface from the IA_PD the WAN interface gets doled
out. 
</hs>

If that's so, since the document's focus is on router behaviour, perhaps
it is this behaviour should be explicitly mentioned.

Section 5, paragraph 1 states 'optional Loopback network interface,
facing the Service Provider upstream, which initiates stateless DHCPv6'.
A loopback interface doesn't really 'face' any direction.  You might
want to change the wording to something along the lines of 'optional
Loopback network interface initiates a DHCPv6 request toward the
upstream SP'

<hs>
I agree with you. My intent was to only mean that this interface
initiates any address acquisition to the SP. Will take your suggestion.
</hs>

Also I notice in the same paragraph, "recommend the CPE Router WAN
interface acquire its global IPv6 address using stateful DHCPv6 for
administrative control of the router".  This makes it sound like DHCP is
recommended for administrative control where I think you mean the global
address is recommended for administrative control and provisioning of
proxied services to the downstream devices.  Why not just say "recommend
the CPE Router acquire a global IPv6 address for administrative control
..."

<hs>
Ah, no we don't mean what you suggested with your text. What we mean by
administrative control is that if one uses DHCPv6 then the address
acquisition by any downstream client is known to the Sap's provisioning
servers like the DHCPv6 server and also a backend provisioning systems
(BPS) that serves multiple DHCPv6 servers. If clients acquire IPv6
address using SLAAC, then only the SP router is aware of these clients -
we will need some protocol to run between the SP aggregation router and
the BPS to inform about clients. Cable IPv6 standards have prohibit a
cable modem to initiate SLAAC and mandate that the cable modem initiate
stateful DHCPv6 for address acquisition. This mandate is similar to the
administrative control we are talking about.
</hs>

Hemant

Antonio Querubin
whois:  AQ7-ARIN