UUNET peering policy

Sometimes I think "Internet Time" also applies to its long term memory.

One of the reasons why I asked for copies of old peering policies is
because essentially every major provider has publically announced their
policy at one time or another in the last decade. Saying "First" about
anything should mean more than the last 18 months.

InternetMCI had their peering policy on their web site for several years
(1995-1996) prior to its acquisition by Worldcom. I believe InternetMCI
removed their peering policy from its public web site about the time Farouq
took over peering at MCI.

The first peering battle was ANS. The agreement was brokered by BBN arranging
for ANS to connect to the CIX router. At that point, the definition of
"being on the Internet" changed from being connected to the NSFNET to being
on the commercial Internet, and the set of providers supplying commercial
Internet service.

The second peering battle was one of packet loss. Sprint tried to make things
as painful as possible by never upgrading its connection to the CIX router
above a T1. So even though other providers were exchanging traffic at 34Mbps
to 45Mbps, Sprint kept their quality of service limited to 1.5Mbps at the CIX.
This one was never directly resolved. But by this time most providers were
exchanging a majority of their traffic via MAE-East.

Around this time BBN transitioned from being a customer of InternetMCI
to being a peer of InternetMCI using its connections via several old
NSF regional networks (BARRNET, SURANET and NEARNET). BBN and MCI may
have had the first "private" circuit peering. Because InternetMCI had
sold connectivity as a loss-leader to the old NSF regionals, some folks
throught MCI was happy to get out of the customer contract.

The third peering battle involved disconnectivity. BBN was one of the first
providers to terminate its connection to the CIX router, which had previously
acted as the peering point of last resort, and began the second round of
peering disputes. When all the major providers connected to the CIX, it was
difficult for any provider not to peer because the CIX router always offered
a way to exchange traffic. In less than three months, BBN, MCI and Sprint
actions eliminated CIX as the router of last resort.

It should be noted, UUNET has maintained its connection to the CIX
router. Any provider interested in exchanging traffic with UUNET has
always had the option of sending traffic via the CIX. This option does
not exist for Genuity or Sprint.

The fourth peering battle involved AGIS announcing its new peering
policy at the least NANOG meeting held at the University of Michigan.
It generated a lot of noise, but eventually AGIS's peering policy
became irrelevant. Towards the end, AGIS was actively trying to get
peering with new providers.

The fifth peering battle involved UUNET. UUNET notified some number,
I've heard between 10 and 20, providers UUNET would terminate their
peering. At this time in the Internet's history only a few providers
had written peering agreements. There were very few NDA's involved
with peering before this time. It probably wasn't a breach of NDA, but
someone leaked the story to the press. UUNET eventually was able to
shutdown the story, but that lead to the next problem. Everything is
a secret, so people imagine things were in peering agreements.

The sixth peering battle involved once again BBN/GTE and the MCI/Worldcom
merger. GTE worked very diligently to bring the issue of peering to the
attention of regulators in the US and Europe. Eventually the EU Commission
issued administrative inquires of all the major providers about the nature
of the agreements, the amount of traffic, the types of connections and
so forth. In the end, the nature of peering agreements wasn't clarified,
but Worldcom had to spin off its InternetMCI division to Cable & Wireless.

The seventh peering battle involved again BBN/Genuity/GTE and Exodus. This
time it was the battle over imbalanced traffic flows. BBN and Exodus had
a dispute, but it was settled and as they say on TV the terms were not
announced. The imbalance issue has come up a few more times with other
providers such as PSI, Abovenet and others.

So, even though some folks like to point to UUNET as the big bully on
the block, if you look at history; BBN has more often than not been
the power behind the throne in these peering battles.

Sean Donelan wrote:

One of the reasons why I asked for copies of old peering policies is
because essentially every major provider has publically announced their
policy at one time or another in the last decade. Saying "First" about
anything should mean more than the last 18 months.

If we want to talk about first, I guess we should acknowledge that the
first transmissions between UCLA, SRI and UCSB in Nov of '69 resulted
from the first peering agreements ;-). But I digress...

Sean, you're right in that there have been many major and minor peering
battles, with lots of "publication" of policies at different times. The
point I attempted to make, apparently unsuccessfully, was that in the
current round, after the EU battle, Genuity posted its policy publicly
to defuse any claims of the pot calling the kettle black in front of the
politicos. And UUNet was effectively and finally forced to do the same
recently. I assume no-one would argue that UUNet is the 800 lb
gorilla... and based on the specifics of the policy, and the fact that
it has been published to attempt to debunk the existence of an "old
boy's network" in peering, they would find it very difficult to refuse
peering arbitrarily. And I believe that this will a) create an
environment and reason for regionals to merge, and b) prove to be a
prelude to a Worldcom/UUNet acquisition binge, maybe including some
"really big" networks.

That's all.

I went back and looked through the records, and found an announcement
from John Sidgemore on May 13, 1997, discussing the change in peering
you mention (http://www.zdnet.com/zdnn/content/inwo/0513/inwo0001.html).
However there was never a public document that I can remember, or
locate. Did you find a public document with _specific_ requirements from
UUNet in the past? My peering agreement outside of CIX with them in 1995
was verbal, with a hand shake. And G-d knows the traffic was unbalanced.

Consider publishing your research and records. And perhaps we can get
others to contribute. It is historic. I kick myself for not having kept
the emails that flowed between the designated contacts of the core group
connected to the CIX router in Santa Clara in the early 90's. The
colorful language and entertaining flame wars between people who have
become rich and famous (and some who have disappeared) were magical in
hindsight. Bob Collett, Marty Schoffstall, Paul Vixie, Rick Adams, and
Bill Washburn were regular folks back then :wink:

sean@donelan.com (Sean Donelan) writes:

...another stunningly well researched, accurate and articulate article,
including the following:

... The imbalance issue has come up a few more times with other
providers such as PSI, Abovenet and others.

To the best of my knowledge, AboveNet has never insisted on any particular
traffic balance with any of our peers. Send to us 10:1, 1:10, 1:1, whatever.
Any traffic coming or going over a peering connection is to or from one of
AboveNet's customers, which means we're paid (by that customer) to deliver it.
(Any other policy amounts to wanting to be paid twice for the same packet.)

Of course I can't commend on PSI or any of the other companies Sean mentioned.

Sun, Jan 14, 2001 at 01:24:32AM -0800, Paul Vixie:

sean@donelan.com (Sean Donelan) writes:

...another stunningly well researched, accurate and articulate article,
including the following:

> ... The imbalance issue has come up a few more times with other
> providers such as PSI, Abovenet and others.

To the best of my knowledge, AboveNet has never insisted on any particular
traffic balance with any of our peers. Send to us 10:1, 1:10, 1:1, whatever.
Any traffic coming or going over a peering connection is to or from one of
AboveNet's customers, which means we're paid (by that customer) to deliver it.
(Any other policy amounts to wanting to be paid twice for the same packet.)

yes! hopefully those who insist upon particular ratios groked that.