[arin-ppml] Draft Policy 2008-3: Community Networks IPv6 Assignment

[Lea Roberts]

On Tue, 24 Mar 2009, Owen DeLong wrote:

> There is a disconnect between original IPv6 marketing hype
> and reality here.

regardless of marketing hype, there are many things that were said at even
the IETF level which have not been delivered in the protocol that exists
today...

> The "simple renumbering" and independent internal addressing
> structure capabilities are not fully baked and have not as yet
> materialized in IPv6.

FWIW, the simple changing of prefixes on host interfaces, by adding a new
router advertisement and deprecating the old, has been demonstrated to
work (mostly).  unfortunately, however, there exists no tie-in from that
mechanism to access-lists and firewall rules and such that still remain
administrator intensive.

> On Mar 24, 2009, at 12:08 PM, Ted Mittelstaedt wrote:
>
> >
> > I had thought that one of the big advantages of IPv6 is that it was
> > designed to be simple to renumber.

certain parts of the renumbering process are simple but still require
several steps even if the exceptions noted above are not involved.  one of
the cool things that works but has yet to have practical uses is that an
interface can have multiple prefixes assigned.  the hard problem there is
how you decide, among multiple prefixes with global scope, which one to
choose as best to initiate a connection off-site...  this is the kind of
thing that the "experimenter" type of community network might be able to
help develop and test.

> > Thus I am not sure why having "a stable and globally unique address
> > assignment" has anything to do with having "a stable internal address
> > structure" under IPv6.  I can understand why a community network would
> > need the second thing, but I don't see why they can't have this under
> > a globally unique address assignment that's made by a LIR instead of
> > by ARIN.

you seem to assume that they would have a stable provider (i.e. LIR)
connection, which the proposers indicated was not the case for many
community networks.  a better argument would be for them to use ULA but
that's not guaranteed to be globally unique.

> > The community network's internal address structure would NOT change
> > when their connections to outside networks come and go - under IPv6.

once they have an assignment, that is true.  but if not from an LIR???

> > Could the proposers explain what they need, here?  We all what to
> > support non-profit community networks that help poor people get
> > online, but at first blush this looks like the proposal authors are
> > assuming IPv6 == IPv4.

I assure you they (and we on the AC) are quite aware of the difference...

since you ask for disclosure:  I am one the AC shepherds for this policy
and I am in favor of this policy.  My co-shepherd is very much against it.
under the new PDP, I've edited the suggestions from other AC members into
the text as posted and rewrote the rationale (obviously not well enough...
:-)  so while I'm not one of the original proposers, I guess it's on me as
more like an author of what you now see.  I'm sorry to have failed to
explain clearly enough why allowing these IPv6 assignments is worthwhile.

frankly, I believe that current policy limiting IPv6 assignments is much
closer to "assuming IPv6 == IPv4" than this proposal.  I already gotten
all the arguments about how "routing slots" are precious, as did the
original proposers.  they understood that getting into the global DFZ was
unlikely but did have hopes of doing some IPv6 experiments with local
peers and perhaps over tunnels.  personally, I don't think the LIR model
maps very well into the community network environment.  they can probably
get by with ULA but it's sad how the Internet's routing problems bleed
over into creating this false address scarcity.  but, hey, that's the
reality of networking in the early 21st century!    regards,  /Lea