Re: New version of the "IPv6 Unicast Address Assignment Considerations" draft

[Mark Smith <ipng@69706e6720323030352d30312d31340a.nosense.org>]

Hi Chip, Everybody,

On Thu, 6 Jul 2006 19:51:40 -0400
"Chip Popoviciu \(cpopovic\)" <cpopovic@cisco.com> wrote:

> Dear v6ops WG members,
>  
> We submitted version -01 of the "IPv6 Unicast Address Assignment
> Considerations" draft today, past the June 26 deadline which means it
> will not be published in time for IETF 66. This version integrates the
> WG feedback and the new Service Provider addressing case study. We will
> review these changes during our presentation in Montreal so we wanted to
> make sure that this new version is available to the work group, even
> though it comes as an attachment to this e-mail.
>  

I've been reading through the SP section, and have some comments.

I realise that the SP network that is used as an example is an
example. However, while reading through it, I seemed to be trying to
read or find a description of a general model for addressing that would
apply to any SP network, rather than an MPLS one. While reading through
the description, I found that I was trying to map those MPLS specific
terms back to more general purpose / functional ones, probably in part
because at the smaller SP I work for, we don't have MPLS deployed yet,
so I don't "think" in MPLS terms yet.

I'd suggest the first section of the SP addressing part could describe
a general model of SP IPv6 addressing. A subsequent section or sections
could describe technology specific deployment models as well as specific
considerations or constraints that addressing causes or effects, such
as the MPLS label conservation concern, or IGP scaling.

Following a model description, I think a section summarising or listing
the addressing space allocations/layout would be useful, matching the
address blocks to their purpose. This would provide a single overview
of the address plan being used.

I'd also think a section describing how to better re-balance the size
allocations would be useful, as well as describing constraints and
consideration of those resizings. For example, at the SP I work for, we
have a lot more smaller customers (residential / small business), and
therefore reserving /46s for the each /48s we'd be likely to give those
customers would be excessive. While I personally generally prefer the
idea of giving all customers a /48, the per-customer /56 that has been
suggested in other forums would likely be plenty of address space for
most (if not all) of our customers, so even /48s can be considered
excessive.

Hope this helps,
Mark.