[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Comment-Aso] IP address space allocation: not a crisis, let's lighten up.

Hi Esther...here's another issue.

ICANN should take care to avoid uncritically adopting the
authoritarian ARIN/RIPE/APNIC model.  The structure of the ASO should
ensure that proposed alternatives to authoritarian allocation are not
strangled at birth by these entrenched organizations.  Moreover,
either ASO or ICANN should *encourage* exploration of the dimensions
of the true problem(s), and exploration of radically improved ways to
solve them.

A friend's ISP demanded that he submit a pile of personal details or
his IP addresses would be taken away.  This policy was claimed to be
forced on the ISP by ARIN.  ARIN is like NSI (which it spun out of):
if you don't like their policies, you merely have to get off the
Internet.  It makes them a wee bit arrogant.

I sent ARIN a note, and got no response.  I wondered whether there is
any need for their harsh policies, which were created when the routing
tables grew too large for the routers of that day to handle.  ARIN was
disinclined to investigate whether or to what extent it needs to
exist; it exists by fiat, like the FCC it resembles.  So I asked kim
claffy of CAIDA for real data.

Kim exhibited a chart of IP address allocation at IETF two years ago.
Hans-Werner Braun has updated the chart recently and also included a
time-series (URLs below).  This is data measured from the real
Internet by real statisticians.  It shows that in the last two years
we've burned up about 3% of the IP address space (going from 21% to
24%).  At that rate, we have at least a few decades before we are "out
of IPv4 addresses" or IPv6 is a critical need.

Many people believe the problem is much worse, because they have been
told so.  E.g. the EuroISPA comment on the ASO proposal says, "The
pool of available addresses is becoming ever scarcer as worldwide
demand rises sharply."  I've been told that the sky is falling so
vehemently that I expected by y2k we'd have burned up half the
addresses -- instead we're only a quarter of the way there, and the
trend is surprisingly flat.  (Seeing data from before 1997 would be
particularly valuable in discerning long term trends.  HWB, kc: do we
have any snapshots of the routing tables from before then?)

Perhaps what's going on is a Y2K-style issue in the making.  If people
don't switch gradually to IPv6 then we will eventually have a crisis
at date X (2010? 2020? 2030?), so let's simulate a crisis now, so the
problem will get solved while the net is still relatively small.

Once we know the problem isn't as bad as it's thought to be, the
policies for who can have how much address space are probably too
harsh.  Requiring detailed information about the users of every subnet
is clearly over the line.  But beyond the minutiae of allocation
policies, the entire concept of administrative allocation should be up
for questioning.  There was an experiment proposed in auctioning off a
chunk of IP address/routing space a few years ago, along the lines of
the early USG spectrum auctions.  I think such experiments should be
tried.  Ultimately, routable IP address space should be allocated by a
real market, as an alternative to kissing the feet of a bureaucrat to
get an allocation.  Such a market would give people very clear signals
about when to invest in IPv6.  The Internet community has been moving
further away from such markets, probably because of a lack of economics
background among Internet architects.

As a third point, networks are being designed for a scarcity of IP
addresses, when the scarcity is somewhat artificial.  E.g. the new
worldwide 3rd-generation digital cellular standard proposals do not
provide IP connectivity from the end-user's handset to the Internet.
They couldn't give an IP address to every handset, the way they can
give it a telephone number that works wherever it goes, so they have
their own protocol which they will "translate" to IP at some set of
protocol translation gateways somewhere.  This is a very unfortunate
development; it means that cellular operators are continuing the
separation between the mobile network and the Internet.  This is
requiring new entrants like Metricom to (slowly) wire the world for
mobile IP.

The Palm Pilot that comes with wireless connectivity is an example of
this model -- you can't actually access the Internet through it, only
a strange subset defined by people who cut deals with Palm.  Cellular
providers would love such a model, which puts them in the catbird seat
granting "yea" or "nay" access to each Internet startup, but a global
network architecture should also permit competitors to come along
and offer the users a choice of real IP connectivity.  Phil Karn
<karn@qualcomm.com> has written some thoughts around this issue.

Address allocation policies were made in a mindset of "crisis", but no
crisis now apparently exists.  There are issues, but no crisis.  The
original crisis came from the size of routing tables, not an
exhaustion of IP addresses.  The regulation that resulted has been
propagated into regulating not only the number of routing table
entries, but also how many IP addresses each such entry covers, who
has them, and what they are used for.  Now there's nothing so
permanent as a temporary emergency, and we have three organizations
dedicated to maintaining that emergency.  It's unlikely to ever end if
ICANN listens solely to them about address allocation policy.


------- Forwarded Message

Date: Mon, 11 Oct 1999 16:15:17 -0700
From: k claffy <kc@caida.org>
To: John Gilmore <gnu@toad.com>
Cc: Paul V <paul@vix.com>, hwb <hwb@caida.org>, sean <mccreary@caida.org>
Subject: Re: John Gilmore: Chart or statistics on IP address space allocation

hwb's done updates
(i still want him to take out the big chunk
in the bottom bitmap on 
cause it's just funky artifact of pv's rbl stuff
but i haven't gotten a chance to nag him enuf yet)

above sentence may make more sense after you read it all...
no promises

i note that hwb is the only one i know
doing anything legitimate to track this 
and there's tons more to do if we had more time/elves,
including analyzing actual traffic traces from a bunch
of places to color by: 
	. delegated
	. allocated
	. announced
	. *used* (i.e., traffic actually going from or to those addresses) 

obviously this last bullet is way sampled
and even so requires special access 
to 'representative' traces
that few have

another cool thing to do would be to color by
those addresses that have reverse in-addr's lit up
(color the rest some punk shade of pink)

you send me elves
i'll send you research

On Thu, Oct 07, 1999 at 12:22:40AM -0700, John Gilmore wrote:
  I never got a response to this from ARIN.  Do you have any info
  on the history of the rate of exhaustion of IPv4?
  ------- Forwarded Message
  To: webmaster@arin.net, gnu
  Subject: Chart or statistics on IP address space allocation
  Date: Sat, 18 Sep 1999 16:53:00 -0700
  From: John Gilmore <gnu@toad.com>
  I looked over your site but don't see any information about how
  well the allocation of IP addresses is going in the long term.
  I.e. what % of IP address space is currently assigned?  What is
  the trend in the global routing tables governed by address assignment
  The ARIN web site lays out a completely authoritarian, almost Germanic
  command structure in which nobody has any rights and everything is at
  the discretion of the ARIN staff -- provided that you submit
  everything in triplicate and with full privacy violation.  I'm
  wondering whether this is actually *necessary* and whether, if claimed
  to be necessary, it is actually *working*.  Have any stats on this to
  justify all the paperwork you put people through?
  (I thought I'd seen a chart at IETF that showed that only about 1/4 of
  the address space is yet allocated, and at the expected rate of
  allocation it is likely to last for decades or more.)
  ------- End of Forwarded Message

------- End of Forwarded Message