Re: [sig-policy] IPv4 countdown policy proposal
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On Feb 16, 2007, at 9:11 PM, Geoff Huston wrote:
At 12:40 PM 17/02/2007, Robert Gray wrote:
Kosuke Ito wrote:
To define the gradual level of what extent we need to reduce
should be the next step of issue, if we found that our current
proposed approach does not provide enough time range
for communities to shift into IPv6.
Ito-san,
At the heart of the matter, my concern with your proposed policy
is that
it does not deal with providing an orderly transition to a world
where
only IPv6 allocations are available.
It always struck me that IPv6 represents an incremental and quite
conservative step in technology over IPv4, but without the
conventional attributes of incremental piecemeal deployment that
technology incrementalism usually achieves . The fact that IPv6
deployment is not incrementally deployable, is not backward
compatible with IPv4, and is not a deployment that makes sense to
undertake in isolation, makes the business case fro deployment
really hard to phrase for many actors. The alternative, of more and
more dense NAT deployment, simply transfers the cost of address
scarcity to others and stays within some form of "comfort zone" of
not changing all that much from where we are today. Its not that
IPv6 is the "only" way forward here, but perhaps the more
constructive question is what is the _preferred_ way forward, and
can the environment be structured so as to make such a preferred
path a 'natural' one for industry actors to follow?
The other question that I've been asking myself in this topic
stream is how and why adoption of this form of policy regarding
IPv4 unallocated pool exhaustion would assist us to transition out
of the current address distribution regime without major negative
forms of disruption to the Internet. Personally, I've not yet seen
a clear and convincing case (for me at any rate) to answer this
concern.
I think it makes sense as a kind of "crisis planning" initiative, and
works in basically the same ways:
1. In case of emergency (inevitable in this case), an orderly
evacuation path is specified. The proximate goal/destination is
"out"; working out more specific post-crisis destination(s) is
another matter entirely. Panic about the event itself should be
reduced, if not eliminated entirely.
2. Publication of the plan reminds people that crises are possible
(inevitable in this case). This may cause some of the panic to occur
earlier -- at time of publication rather than at the time of the
emergency itself -- but even so spreading the disruption out over a
longer duration, as opposed to dealing with everyone's all at once,
should in principle make it easier for everyone to manage.
3. Early responders may create greater demand for more constructive
transition mechanisms, sooner rather than later. With luck, an
earlier start at exploring potential successor arrangements will help
to reveal promising vs. problematic alternatives, sooner rather than
later.
On this view, the proposed mechanisms tend to optimize for near-term
order and stability, although potentially at the expense of long-term
order and stability. Unless (even more) 11th hour panic would
actually be constructive, or the preferred destination is well known
and a better transition can be formulated to lead more directly to
that successor state, this may be the best we can hope for at the
moment...
TV
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)
iD8DBQFF1nATUHTO4sHEFsERAuDHAJ42ouDGtPyYN3M4skWNfousHbjXogCbBTtM
r1C3lX+uS0R0mwv05jI/KzU=
=mOlc
-----END PGP SIGNATURE-----