[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Posted a new copy of CPE Rtr draft
On Mon, 16 Mar 2009 11:22:42 -0700
Fred Baker <fred@cisco.com> wrote:
>
> On Mar 16, 2009, at 5:55 AM, Maglione Roberta wrote:
>
> > could you please explain why stateful DHCPv6 is the recommended
> > method for assigning a global IPv6 address to the WAN interface?
> > Are you suggesting excluding the possibility to use SLAAC to number
> > WAN interface?
>
> I don't understand him to be *precluding* anything. However, I do
> believe that he is saying that an ISP generally has more to configure
> on a CPE router than the interface address; it will have the
> recommended DNS server, the prefix that the SOHO/Residence will subnet
> internally, if DS-Lite is in use it will have the necessary
> information to configure a tunnel, and so on. He is saying that if
> that is the case, one may as well get the interface address from the
> DHCP server as well.
>
> I think the question before the house is whether the ISPs, such as
> Telecom Italia, agree with that viewpoint. Certainly you can *use*
> SLAAC in your network. Would you rather the specification *recommend*
> SLAAC? If so, do you have a recommendation regarding the other
> matters? Would you prefer that the document simply recommend that CPE
> equipment support both options?
>
> I am not expressing an opinion here; I know of networks in which DHCP
> seems like a better choice and networks in which SLAAC is a perfectly
> reasonable solution. I think it is important that you - and other ISPs
> - state their opinions however.
>
(not representing a ISP here, but work at one, and these
sorts of issues are some of the ones I have to worry about)
I think DHCP is the better option. The link to the customer's WAN
interface is the first customer unique or specific part of the
network connection, so having some non-upstream-router harder or hard
state in your management systems i.e. DHCP database, can make
troubleshooting and billing easier. It also encourages the model that
ISPs give out subnets to all customers (unless there is a PD option for
RAs that I'm not aware of), not end-node addresses.
Regards,
Mark.