[arin-ppml] IPv6 Non-connected networks

[Michael Richardson]

>>>>> "David" == David Farmer <farmer at umn.edu> writes:
    David> Since I've been working on rewriting ARIN's IPv6 policy I
    David> want to know how the community want this to actually work, my
    David> goal is a more direct and clear path for these network to get
    David> addresses.  However, I see a faction of the ARIN community
    David> that want these assignments made from a filterable block.
    David> Then, there is another faction that says, wait a minute, that
    David> means ARIN is defining Routing Policy.

If the assignments are not made from a trivially filterable block, then
the assignments are not special in any way, and if you can convince an
ISP to carry them, they are exactly the same as connected blocks, and so
should be charged in the same way.

The argument against giving out IPv6 /32s to any and all (regardless of
need) is, I think, that it causes DFZ router table explosion.

(Remember, there are enough /40s to give every man, woman and child one,
but I see no reason why any individual human needs more than a /60...)

We are told that ARIN doesn't do routing policy, yet it seems to me that
a large amount of ARINs reluctance to make IPv6 addressing available is
the result of a concern about DFZ router size.

If we had a way to assign a routing table congestion cost to
assignments then maybe it would be okay.
Maybe SIDR will provide a mechanism to do that, since the "routability"
of a block could expressed in the authorization certificate binding it
to an ASN.  

Until such time, ISPs are guardians of a scarce public (distributed)
resource known as the DFZ.  It's a tragedy of the commons, and a
fax-effect (at the same time),  were once one ISP starts to pollute the
table, other ISPs are encouraged by peer pressure to do the same thing.

    David> Basically it boils down to, if we assign from a common pool
    David> then there is no distinction and you can only have the
    David> address space you can justify connected or non-connected.  If
    David> we make a distinction then, anyone must to be able to get
    David> such a non-connected assignment without regard to if they
    David> have an assignment intended intended for global connectivity.

I can't parse the second sentence, but I think I know what you mean.
Can you rewrite it for me?

    David> So I would like to see continued discussion on the specific
    David> issue of a common pool for all assignments VS. a specific
    David> pool for non-connected networks.  Additionally, I intend to
    David> raise this issue on the floor in Toronto.

Thanks for bring us back to the topic.

(It seems amazing that ARIN has only ever once provided aN IPv4 prefix
that wasn't routable, which is the 24.y/11... )

-- 
]       He who is tired of Weird Al is tired of life!           |  firewalls  [
]   Michael Richardson, Sandelman Software Works, Ottawa, ON    |net architect[
] mcr at 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.