[ppml] IPv6 & /48 [was 2005-1:Business Need for PI Assignments]

[Lea Roberts]

Leo (et al) -

I really agree with you on this.  Let's start a different thread (hence
the subject change...  :-) so the 2005-1 thread can maintain focus on the
PI question.  BTW, I believe that non-PI addresses are part of the same
set of recommendations and so it is time to start defining the policy for
what a reasonable set of constraints to allow PI assignments should be.

The presentations during the IPv6 roundtable on Wednesday morning made a
convincing case for revisiting the /48 gospel.  For those of you who were
not there, check out the first set of slides "Where did all those IPv6
addresses go?  David Conrad channeling Geoff Huston" on the ARIN web site.
By continuing to pass out /48s like candy, we can dig ourselves a real big
hole!  Granted, I won't be around to suffer, but Bill Darte spoke
eloquently about the legacy we leave behind.  Given the pain of changing
the network that we are already experiencing, pretending addresses are an
infinte resource seems very foolish.  Sure the numbers are huge, but as so
many others have pointed out, it's really hard to predict the future.  If
we don't ruin it, IPv6 can live for a long time and perhaps deliver on
some of the yet-to-be-real design goals (e.g. renumbering can be much
easier if the needed tools for DNS and ACLs are built... ;-)

In any case, I find it easy to say that instead of assigning a /48 to home
uesers, instead assigning a /56 to home users is an easy way for being
less wasteful with being overly restrictive.  If someone has a big enough
home to need more than 200 subnets, then they can have their /48, but your
"normal" home user couldn't possibly be overly constrained by that
assignment.  One of the comments during the discussion of the IPv6
roundtable pointed out that the /64 is also likely to be too little for
the cell phones of the future which may well be the mobile router for your
car network...  :-0  so maybe cell phones will start getting /60s instead
of /64s.

In any case, let's try to get away from the belief that there are so many
addresses we don't have to even think about address consumption and let
the pendulum swing a little bit back toward the center.

							/Lea
On Mon, 25 Apr 2005, Leo Bicknell wrote:

> In a message written on Mon, Apr 25, 2005 at 11:48:13AM -0700, Owen DeLong wrote:
> > While the /48 and /64 decisions could be policy, given the current
> > constraints
> > and assertions in the V6 protocol specifications, it is not a policy
> > decision,
>
> The /64 is a design decision consequence of RFC 2462.  However, the
> /48 boudary is a recomendation in every draft I can find.  Indeed,
> RFC 3177, the most on point here, starts right off with "this
> document provides recommendations to the addressing registries",
> and then proceeds to defend those recommendations.  RFC's 3513 and
> RFC 3587 don't make any further requirements.
>
> So, while it may take the IETF to fix the /64 mess, the /48 mess
> can be fixed in RIR policy, including steps like moving it to a /56
> or if you were so inclined a /44.
>
> If someone can point to where the IETF _requires_ a /48, I'm all
> ears as I've looked and can only find recomendations.
>
> --
>        Leo Bicknell - bicknell at ufp.org - CCIE 3440
>         PGP keys at http://www.ufp.org/~bicknell/
> Read TMBG List - tmbg-list-request at tmbg.org, www.tmbg.org
>