RE: I-D Action:draft-ietf-v6ops-ipv6-cpe-router-03.txt

[<Jason.Weil@cox.com>]

In regards to the sentence in WPD-1: 
"The IPv6 CE router MUST ask for a prefix large enough to cover all of its LAN interfaces."

I believe this sentence is useful for the following reasons:

-Assure contiguous PD requests in a basic CE Router environment (single-homed, no sub-PD) 
-Minimize prefix injection - syncs prefix injection on the SP access router to the prefix size requested by the CE Router (this may or may not be helpful depending on the allocation method but could useful in certain situations)   

Like Barbara mentioned many service providers may choose to control the response from the delegating router to a certain prefix length but this is not assured and may not be implemented by every SP. For basic single-homed operation in an efficient manner the CE Router would determine the number of valid LAN interfaces and ask for an aggregate prefix that covers that range. One option here that would be valid without this statement would be for a CE Router to ask for a separate IA_PD /64 for each interface in separate messages which would be less than optimal.   

Note this by no means limits the CE Router from requesting unique IA_PDs when multiple prefixes per LAN are required and as stated previously prefix subdelegation is out of scope for this document. 

Thanks,

Jason


________________________________________
From: owner-v6ops@ops.ietf.org [owner-v6ops@ops.ietf.org] On Behalf Of Brian E Carpenter [brian.e.carpenter@gmail.com]
Sent: Saturday, December 19, 2009 2:58 PM
To: Mark Smith
Cc: IPv6 Operations
Subject: Re: I-D Action:draft-ietf-v6ops-ipv6-cpe-router-03.txt

On 2009-12-19 18:42, Mark Smith wrote:
> On Fri, 18 Dec 2009 09:09:21 -0800
> Fred Baker <fred@cisco.com> wrote:
>
>> I will open a WGLC on this after new years; My mind will be elsewhere
>> for the coming two weeks, I imagine yours will as well. However, if
>> you want to start reading/commenting now...
>>
>
> Maybe I'm being a bit overly paranoid about people being precious with
> IPv6 address space, however, the last sentence of -
>
>
> "WPD-1:  The IPv6 CE router MUST support DHCP prefix delegation
>            requesting router behavior as specified in [RFC3633] (IA_PD
>            option).  The IPv6 CE router MUST ask for a prefix large
>            enough to cover all of its LAN interfaces."
>
> could be interpreted to mean that the ISP only needs to provide a
> prefix to meet the requested size e.g. for two LAN interfaces a /63.
> That would seem to me to be encouraging an "only enough" address space
> model, rather than a "more than enough" address space model, which is
> what I think IPv6 is aiming at.

Without disagreeing with anything Barbara Stark said, I'm wondering
why the last sentence of WPD-1 is there at all. I mean, why would
an implementer ever ask for too little address space?

>
> There might be some value in stating that it is likely that an ISP will
> delegate a prefix that not only meets this minimum requirement but also
> allows for a number of additional subnets downstream of the CPE, with
> the delegated prefix size likely significantly larger, possibly a /48
> or /56. IOW, stating that the delegated prefix will likely be larger
> than what is requested (not that it is probably likely, but I could see
> a CPE vendor adding in a check to see if the delegated prefix size was
> equal to that requested, and if it didn't, not using it).

So if there is anything to say, it's that the CE MUST ask for at least
one /64 per interface and MUST be configurable to ask for more.

    Brian