Re: [sig-policy] prop-091: Limiting of final /8 policy to specific /9
On Jan 24, 2011, at 5:58 PM, David Woodgate wrote:
>
> "Necessity is the mother of invention" - old adage
>
> Basic economics has always indicated that there would be industry no
> change from v4 to v6 until it was necessary, and the only real
> necessity for doing so is the exhaustion of IPv4 addresses. So I
> don't think that the lack of industry deployment of v6 until now
> should be a surprise, and neither does it mean it has "failed" - I
> believe it has been simply a case of waiting for the right time, and
> that time should be from the RIR exhaustion over the next 12 months.
>
> Nothing would drive IPv6 adoption faster than useful content only
> being available via v6, which would force customers to demand their
> (internet and application) providers make that content available
> somehow, driving the providers towards v6.
> Of course, it is too soon right now to make content IPv6-only,
> because an insignificant number of people would be able to access it,
> and therefore its availability could never be known by the general
> Internet community.
>
It would, however, be interesting to see some leading content provider
get brave and announce the date (some time in the future) by which
they intend to turn off IPv4 (at least for some period of time).
Imagine, for example if $SOCIAL_NETWORK were to announce
that effective January 1, 2013, they would be turning off IPv4
access to their services for a period of 2 weeks. Users that
do not yet have IPv6 (click here to see if you have IPv6 yet) should
contact their ISP to arrange for this upgrade.
Unfortunately, the reality of economics also says that it's unlikely
any $CONTENT_PROVIDER will get that brave until there is a
much more significant IPv6 deployed base. Heck, there are some
that are still hesitating to turn on IPv6 access to the general public
because they might have a problem with as much as 0.05% of
users.
> But if content providers are able for many years yet to get enough
> small amounts of IPv4 addresses to make their content available via
> IPv4 (even if dual-stacked with v6), then this risks removing much of
> the industry driver towards IPv6 adoption, and that scenario would be
> likely to create artificial commerce, black markets and exorbitant
> effective pricing in v4 addresses, and would favour providers with
> well-established v4 networks.
>
I think it is the eyeballs that will be motivated to move first, unfortunately.
Content would be a great motivator for the eyeballs, but, content
isn't motivated to move until the eyeballs do. (Which is unfortunate
because if content had moved first, we'd be facing a much smoother
transition).
However, the eyeball networks simply can't sustain their growth
models and services without an influx of new addresses on a
continuing basis. Those addresses in a market for IPv4 will be
cost prohibitive compared to the economics of residential access.
So, eyeball providers will be forced to deploy a combination of
IPv6 and NAT444. The end result of this deployment will be a
situation where content providers will have users complaining
of degraded user experiences over NAT444 until they implement
IPv6. This will drive the content providers to IPv6.
This is the rough road we have ahead of us. The more content
providers that get in front of this NAT444 deployment with IPv6
sooner, the more we can even out the surface of the road.
Of course there is also the potential issue that some access
providers will deploy NAT444 without bothering to deploy
IPv6 to their customers. I'm hoping that scenario will be
rare enough that customers just move to the ISPs that offer
IPv6. However, even in that case, there will be tunneled
access to IPv6 solutions available to those users which will
still make content accessible over IPv6 a better alternative
to IPv4 content via NAT444.
> So this is another issue I have with the current final /8 policy; I
> think it's a reasonable idea to have enough addresses for specific
> purposes for a couple of years while the transition to v6 gets under
> way properly, but if either it promotes a lengthy availability of
> IPv4 for new content servers (thus potentially removing IPv6 adoption
> drivers and increasing IPv4 address value), or it causes an explosion
> of requests artificially tailored to bypass the intent of the policy
> (thus potentially causing an inappropriate distribution of addresses
> anyway, with unnecessary hostmaster workload and impacts upon routing
> tables, etc.), then I feel it may do more harm than good for the industry.
>
Yep.
> So the question in this context is what the correct balance in size
> of space to reserve? The current policy suggests a /8; prop-091
> suggests another view.
>
With statistics to back up that view. The alternative to make the
existing policy palatable from a statistical perspective would be
to change the limit size to /21 instead of /22.
Owen