[ppml] Policy Proposal 2005-8: Proposal to amend ARIN IPv6 assignment and utilisation requirement - Last Call
Lea Roberts
lea.roberts at stanford.edu
Fri Apr 14 20:30:45 EDT 2006
hi David -
On Fri, 14 Apr 2006, David Williamson wrote:
> On Fri, Apr 14, 2006 at 04:35:25PM -0400, Scott Leibrand wrote:
> > I do agree that home users may want to run more than one subnet, and may
> > want more than one /64. Therefore I agree with the guideline that /64's
> > should only be given out when it is known a priori that one and only one
> > subnet is needed, and that a /56 should be given out otherwise. However,
> > I think that anyone who actually needs a /48 really should be considered a
> > business customer, not a residential customer.
>
> I have to admint that I'm very puzzled by the ongoing effort to keep
> masks bounded by "interesting" numbers of bits.
as I said in the presentation at ARIN XVII, this policy was conceived to
provide for assignments of less than the /48 which was specified by
RFC3177. Geoff's analysis says that, contrary to the belief of the IETF,
assigning /48s to all could lead to premature exhaustion of the IPv6
address space. Moving to /56 for small sites was proposed as sufficient
to protect against that problem. ISP feedback was it was not for the RIRs
to decree what assignment policy an ISP/LIR should have, so the policy was
updated in that respect, with some "suggestions" as to assignment sizes.
The interesting bit boundaries are nibbles (4-bit multiples) since they
define the delegation boudaries for IPv6 reverse delegation. As Owen has
stated in another post, it seems reasonable to avoid the added
complication of not keeping that clean this time around.
> While the vast majority of home users will be fine with a /64, how many
> of the remainder really will ever need a /56? I suspect that the
> majority not sufficiently serviced by a /64 would be fine with a /63 or
> /62. Why assign all of that extra space? It seems to me that you can
> fit a lot of /62s into a single /56. (Am I just missing something
> obvious here?)
there is a real need to encourage generous assignment sizes. there is no
shortage of IPv6 addresses for centuries as long as the shift down to /56
is accomplished. it is much more important to make sure that users have
room to grow as no doubt there will be ways of using addressing in the
future, allowed by IPv6, that we can't conceive of in today's technology.
this policy is just trying to allow for less wasteful generosity than
RFC3177. it is also trying to continue to promote the mind shift away
from the conservative assignment policies required by the limited address
space available in IPv4.
as we also discussed at the ARIN XVII meeting, it would be useful for
some group to define guidelines for assignment policy that would clarify
the issues you raise. it seems that in ARIN policy is not the correct
place yet no other group comes to mind. anyway, as a rough suggestion, I
would say that end sites should get 4 to 8 times as much address space
assigned as they think they might use using today's networking techniques.
I hope that gives you some answers... :-) /Lea
More information about the ARIN-PPML
mailing list