Re: [homenet] draft-gmann-homenet-relay-autoconf-

Michael Richardson <mcr@sandelman.ca> Tue, 08 May 2012 14:30 UTC

From: Michael Richardson <mcr@sandelman.ca>
To: homenet@ietf.org
In-Reply-To: <4FA8C1FB.4080804@gmail.com>
References: <17069.1336418835@marajade.sandelman.ca> <4FA8C1FB.4080804@gmail.com>
MIME-Version: 1.0
Content-Type: multipart/signed; boundary="=-=-="; micalg="pgp-sha1"; protocol="application/pgp-signature"
Date: Tue, 08 May 2012 08:03:53 -0400
Message-ID: <8216.1336478633@marajade.sandelman.ca>
Sender: mcr@sandelman.ca
Subject: Re: [homenet] draft-gmann-homenet-relay-autoconf-
Precedence: list

>>>>> "Brian" == Brian E Carpenter <Brian> writes:
    >> I read this document just now.  What I understand is that
    >> level-2+ routers go to "the" ISP to get an additional /64.  The
    >> ISP could return anything... it might be good if it returned a
    >> /64 adjacent to those already assigned, but it can do anything.
    >> The ISP->customers assignments are not aggregated.

    Brian> Not *necessarily* aggregated. I would have thought
    Brian> aggregation remains valuable.

Let me say:
    The ISP->customers assignments are not aggregated by the customer.

If the ISP aggregates it, great.

    >> This has interesting properties for efficient use of address
    >> space, and also for permitting home networks to be as big or
    >> small as desired.

    Brian> Efficiency at the level of filling /56s really isn't a worry,
    Brian> is it?  It would just be simpler to allocate each subscriber
    Brian> a /56.

I agree, but it's now up to the ISP.

    >> There is really no additional overhead in the homenet, as one is
    >> likely throwing /64 routes around most of the time anyway.  For
    >> the ISP, it's up to them.
    >> 
    >> My only problem is that I don't understand how it works when you
    >> have two ISPs.

    Brian> I think both of these assumptions will be false for a
    Brian> substantial fraction of subscribers:

    Brian>    This document assumes that the vast majority of home
    Brian> networks will connect to a single ISP and will be generally
    Brian> constructed in a tree architecture.

I agree; it's not a good assumption.

I'd rather see the DHCP PD be proxied (rather than snooped) by the
border router(s).  I think that draft-baker-homenet-prefix-assignment
basically does this, but manages the prefixes locally.

So, if we agreed that having the ISP delegate /64s as needed was the
right answer, then draft-baker-homenet-prefix would solve the problem.

-- 
]       He who is tired of Weird Al is tired of life!           |  firewalls  [
]   Michael Richardson, Sandelman Software Works, Ottawa, ON    |net architect[
] mcr@sandelman.ottawa.on.ca http://www.sandelman.ottawa.on.ca/ |device driver[
   Kyoto Plus: watch the video <http://www.youtube.com/watch?v=kzx1ycLXQSE>
	               then sign the petition.

[homenet] draft-gmann-homenet-relay-autoconf- Michael Richardson
Re: [homenet] draft-gmann-homenet-relay-autoconf- Brian E Carpenter
Re: [homenet] draft-gmann-homenet-relay-autoconf- Michael Richardson