[IP] more on China Builds a Better Internet
Begin forwarded message:
From: tomvest@xxxxxxxxxxx
Date: September 26, 2006 10:41:36 AM EDT
To: Fred Baker <fred@xxxxxxxxx>, ip@xxxxxxxxxxxxxx
Cc: David Farber <dave@xxxxxxxxxx>
Subject: Re: [IP] more on China Builds a Better Internet
On Sep 26, 2006, at 8:13 AM, David Farber wrote:
I do wish people would stop propagating that "MIT has more IP
addresses than all of china" meme.
To my shame, I'm the guy who originated that. When I showed the slide
at ChinaInet 2000 that asserted that (and got the name of the school
wrong), it was true. But my point was not that APNIC had committed
some form of wrong-doing. That slide was one in a series that argued
that China, where I was giving the talk, needed to get serious about
Internet deployment, and that specifically to address the needs of a
population of 1.5 billion people it would have to move in the
direction of an IP address space with that many addresses - as
adjusted by the H ratio. They needed to move into IPv6 as IPv4 would
not have the address space they needed.
I have seen that slide in talks selling my competitor's product,
slides selling the ITU as an even broker in the Internet Governance
debate (which I will believe after I find +886 in the enum database;
geo-politics is keeping a recognized PSTN country code out of the
enum database because the relevant communication region's status as a
country is disputed and they are not a member of the ITU), and
others. In most cases, I wonder whether they even know the origin of
the slide.
Of course, at this point China has quite a few more addresses than it
did then. But the entire unallocated IPv4 address space doesn't have
enough addresses to represent the Chinese educational system, much
less China as a whole. The original argument is still valid.
Hi Fred,
Thanks for clarifying the source of this pernicious comparison, and
the rationale. It seems to me that demographics alone makes China's
-- and humanity's -- implicit future requirement for expanded address
resources indisputable. However, the manifest requirement in 2000
(based on ITU stats and relevant firsthand experience) was still
pretty modest -- well below the HD-deflated total address space
allocated to CN by the end of 2000, or the HD-deflated total address
space originated by the 28 publicly visible Chinese ASes that were in
production back then. As I understand it, CNNIC defines an "Internet
user" as anyone who gets online by any means for at least one hour
per month, and there really was (is) no competition or other macro-
level institutional factors that might inflate the addressing
requirement at that time much beyond what one might estimate to be
peak simultaneous usage rates for 9m asynchronous (metered dial-up)
users, 20m ad-hoc "users", 5k DSL subs, and 70k CN-flagged hosts
(most of which, for reasons unique to China, were actually likely to
be there rather than elsewhere).
As you rightly note, China accounts for much more (i.e., 600% +)
address space now -- and still enjoys a comfortable padding between
the current HD-deflated total supply of addresses and the estimated,
peak-sim-usage adjusted total need. China's official interconnecting
carriers are all "very large" members of APNIC, meaning that the only
constraint they face with respect to IP addressing is demonstrated
need, calculated using the same rules that apply to all network
operators these days. CNNIC controls all other domestic IP address
uses, so -- apart from "objectively demonstrated need" at the
national level -- all of the factors affecting supply and demand for
address resources in China are internal to China.
All of this is completely consistent with your recent note, but given
the misunderstanding and misappropriation of the first comparison, I
thought some additional details might be useful. MIT may have been
fortunate in its proximity to the invention of the Internet, much as
other countries/institutions are favored by their "lucky" proximity
to other scarce resources. That said, the extra-national institutions
for IP address distribution that have been in place for the last
decade-plus appear to have done quite a good job of servicing
concurrent "objective" demand, which means that IPv4 addresses should
last as long as possible to satisfy "objective" demand in the years
to come.
Too bad perhaps that the system wasn't established a decade earlier.
Too bad perhaps that it doesn't always assure the same measure of
efficiency across other levels of the address distribution system. We
can only hope that these insights will be preserved, and
strengthened, in the yet-to-be-defined institutions that will inform
the IPv6 distribution process...
Tom
-------------------------------------
You are subscribed as roessler@xxxxxxxxxxxxxxxxxx
To manage your subscription, go to
http://v2.listbox.com/member/?listname=ip
Archives at: http://www.interesting-people.org/archives/interesting-people/