OT - Vint Cerf joins Google

It seems that the list doesn't like the attachments, anyway, the text show
the results for the awstats.


Sorry the late answer, traveling and overbooked ... My reply below, in-line.


De: Paul G <paul@rusko.us>
Responder a: <owner-nanog@merit.edu>
Fecha: Mon, 12 Sep 2005 00:54:25 -0400
Para: <nanog@merit.edu>
Asunto: Re: OT - Vint Cerf joins Google

From: "JORDI PALET MARTINEZ" <jordi.palet@consulintel.es>
To: <nanog@nanog.org>
Sent: Monday, September 12, 2005 12:30 AM
Subject: Re: OT - Vint Cerf joins Google

The last figure that I remember, very impressive, was in April 2004, when
the estimated number of hosts using 6to4 on Windows hosts was calculated


100.000.000 (extrapolated from measurements). This is not including hosts
with have native support or use other transition mechanism such as
configured tunnels, ISATAP, 6over4, or Teredo (behind NAT).

this figure seems to be completely over the top. i would be interested in
seeing those 'measurements', an explanation of why they are statistically
representative and the method of extrapolation. perhaps it was a typo and,
instead of 'extrapolation', they really meant 'exaggeration'? that would
make more sense ;]

The paper is here:

I also know the author, and I'm sure is not exaggerating.

We notice in our web servers (which are dual stack), incredible amounts of
IPv6 traffic, increasing month by month.

please define incredible using a non-subjective measurement system -
absolute counts and percentages of total traffic will do. as stated above, i
would likewise be interested in knowing how representative your traffic is
of general internet usage. as an example, i would expect web servers for an
incredibly popular site discussing v6 to have a disproportionate amount of
v6 traffic.

Just look at the attached stats from this year in one of our web sites. Just
one, and not the one which has the bigger ratio of IPv6 vs. IPv4 traffic. Is
not an IPv6 site, just one of our customers. Can't say how much
representative is vs. Internet traffic, but for me is enough.

The file total.tiff includes ALL the traffic to the server (IPv4 and IPv6),
while the other one (ipv6_only) is just IPv6 traffic. If you compare what is
only IPv4 (total-IPv6) vs. IPv6, we have:

        IPv4 IPv6 %
Users 118.41 GB 10.38 GB 8.77
Robots 253.65 GB 2.64 GB 1.04

Conclusion: The users traffic is rising. No robots support IPv6 today
(probably this could change with people like Google and others doing IPv6).

Different conclusions can be extracted looking at the number of visitors,
visits, pages, hits, etc.

Is clear, that this depends on the user profile, may be even the region ? In
some regions the awareness has been much stronger (and probably successful)
and more users turn on IPv6 in their clients.

Do you want to guess what will happen with Vista, which comes with IPv6
enabled by default ?

i don't like guessing, but if i were pressed, drunk or otherwise
intoxicated, i'd say default support in client software is not the single
bottleneck - being able to purchase v6 transit and have your v6 work as well
as your v4 is another one that you can't really get around. i'm not up to
date on these things, has someone figured out how we're multihoming with v6
yet and, more importantly, got vendors to agree on and implement it?

I disagree here. If the clients have IPv6 support, even if tunneled, which
is enabled most of the time automatically (6to4, Teredo, others), the
traffic is already increasing specially peer to peer. Of course, the quality
is not so good as having native support, but some times it works much better
that having to trouble with NAT boxes and so on.

It seems that the list doesn't like the attachments, anyway, the text show
the results for the awstats.

I believe that the same paper can be obtained for free from

Marshall Eubanks

P.S. I may have inadvertantly sent a blank reply to the list just before
this. (My web based email hiccuped.) If so, I apologize.