[arin-ppml] IPv6 Heretic thoughts
Lee at dilkie.com
Sat Sep 6 11:12:53 EDT 2008
Eliot Lear wrote:
> Lee Dilkie wrote:
>> I don't think the analogy is at all correct.
>> Radio spectrum has a physical hard limit and there is no substitution
>> (until subspace radio is invented). So obviously the effort has *always*
>> been to improve utilization of a fixed resource. I see no reason why
>> this won't continue.
> I don't believe we can say that IPv6 is a substitutable resource for
> IPv4 at this point in time. IPv6 addresses gain you access to very few
> destinations today.
As I said, the analogy is somewhat flawed. However, the move from HF to
VHF could be considered similar. In the early days, when the HF band was
filling up, the only options were to squeeze everyone closer together in
HF (improve utilization) or move to the unused VHF spectrum. where you
had to pioneer new radio technology (somewhat like our IPv6) and deal
with totally different radio propagation characteristics (no analogue to
The same process has been repeated many times, the move to UHF and
higher. But analogy breaks down in several ways. First, moving to a new
spectrum affects your radio propagation characteristics (HF can
communicate over long distances, beyond the horizon, VHF and above is
line-of-sight, HF has limited bandwidth, VHF has much more). We don't
have this issue, IPv6 runs on the same physical medium as IPv4, with no
more or less bandwidth available to it. Second, radio applications often
tie applications to modulation techniques, essentially blurring the
lines between layers, as the band they operate in determines the
bandwidth that can be used. For example, slow-scan TV on HF, fast-scan
TV on VHF (and UHF) and high-def TV on V/UHF (although hi-def also
benefits from the efficiency gained in compression so can operate in
spectrum that technically wouldn't support uncompressed). For the most
part, our networked applications are very much isolated from networking
transport details (IPv4 vs 6). So we *should* be able to move networked
applications to IPv6 fairly easily.
But all that really beside the point.
IPv6 is moving forward, applications are slowing being written to
support it, networks are slowing upgrading. It will be a viable
alternative to the other effort, namely more efficient utilization of
existing IPv4 space.
None of this is going to happen overnight, just as the internet won't
come to a crashing halt after the last IPv4 address is given out. This
will all work out just fine in the end, and frankly, we're lucky as hell
that all our existing networked applications have such an easy
transition as they do to migrate to IPv6. Plus, we're even more blessed
that dual-stack exists and we can network into both address spaces at
the same time from the same host.
In the words of Douglas Adams, "don't panic".
It is largely for these reasons that I don't support a transfer policy.
For one, I'm not all convinced that the radio spectrum auctions have
actually proven to be good. Sure, the companies with the most bucks won
and the government got billions, but some of the deployments have not
been so smooth. In some cases the cost of buying the spectrum has
hampered the business case for the network, either by driving up the end
consumer costs (after all, it's the end user who pays for all business
costs) so much that demand isn't there, or by taking too much capital
out of the infrastructure deployment. Either way, I don't see a
liberalized transfer policy as benefiting anyone but big business at the
expense of everyone else.
To put it simply. I hate derivative markets. If you want to run a public
network, purchase your ip addresses from ARIN at a fixed price (for
everyone), don't start buying and selling addresses on the "open"
market. Because what I worry will end up happening is a market with
folks that hold addresses, to buy and sell, but don't operate networks.
More information about the ARIN-PPML