Sprint peering policy

Mon Jul 1 21:43:35 UTC 2002

> Agreed completely.
> BUT, your logic is very much "perfect world". There are a quite few
> reasons why this doesn't work "in the real world", same as communism
> works great in theory, but not in the real world.

actually, i'd argue that communism's theoretical basis has been roughed up
over the years and that it's pretty much provably stupid even without trying
to build any more existence proofs.  but we're afield.

> Eg:
> #1 Do you honestly believe that you wont run into any customers who will
> say "why should I buy from abovenet if I can peer with them?  They will
> take a big percentage of my traffic and do it for free."  *IF* you could
> have a setup where you could peer AND buy at the same time, then your
> model works better.

there is no useful endgame with a small number of large players.  not only
has wcom proved it, but the fact that only artificial barriers could prevent
new large and small players from increasing the number.  so let's assume a
world where everybody who wants to profit from carrying (bits X distance) is
able to try, and some of them figure out a way to charge more than their costs
(even if those costs include some access fees or recip billing or settlements
with networks larger than them (that year)).

in that model, any enterprise large enough to install in lots of peering
points and buy lots of longhaul will be able to achieve 100% peering coverage
and not have to pay anybody any transit or access fees.  note that these are
huge numbers by everyday standards, and that the longhaul they're buying will
be profitable for some of the dual-stack suppliers (who would also like to
sell IP transit but will be satisfied with the wavelength revenue.)  this is,
as far as i'm concerned, a realistic world-view for fortune 100 companies.

but it will certainly open the door to competition, since any well-enough-
funded IP carrier could build that same network and charge "cost plus"
pricing for IP transit.  i can envision some consortiums sprouting up, with
likely an industry-specific flavour (medical, car production, transportation,
etc) to try to control supply line costs across specific industries.  anyone
with enough manna to build a global footprint would be able to avoid paying
any kind of transit fees or other blood money to anybody else, even while
they were paying for all kinds of capex and salary and wavelength.  sounds
like a healthy, vibrant, competitive environment to me.

but is it utopian?  not according to me.  (witness my support for PAIX.)  all
it will take is for the so-called "tier-1's" to continue winnowing
themselves down as their quasi-predatory closed peering model leaves them
with nobody left to be predatory *toward*.  i'm not arguing that some gov't
needs to step in and fix anything.  the market can self-correct -- like now.
but i will continue to be astounded that the IP market is being _allowed_
to seek its own level, even at demonstrated great risk to consumers everywhere.

there is in other words room for a moderate number of moderately sized
companies to make a moderate profit selling the service of "we will build
a global network and peer everywhere so that you won't have to."  it's when
we cut that number to a dozen, or a half dozen, or just one or two players,
that the tree gets very brittle and the first strong breeze knocks it down.
(no amount of predatory peering practice will keep another Level(3) from
being created and well-funded whenever the existing players are making lots
of cash... even if they aren't keeping (or counting :-)) that cash much.)

> #2 When something is being done for free, it is often not being done as
> well as if it were paid for, case in point: AboveNet's link to NY-IIX is
> 100mbit, right?  The MRTG graphs seem to indicate that it is, AND that
> it was doing 90 megabit for several months straight.  Shouldn't this
> have been upgraded to gig-e?  Not cost effective?
> 
> That being said, I like what above net does, and what they stand for, I
> just don't see how it can possibly make money.

well, so, i'm not MFN's CTO any more, and so i can't speak to their status
since i'm just not in the loop.  but the peering contract used to sign
called for upgrades which a company in chapter 11 bankruptcy can ill
afford, and so i can sort of imagine why they might not have upgraded that
connection lately.

BUT: the open peering model dave rand built abovenet upon and which i
followed during my tenure there was in no way responsible for MFN's later
woes.  some customers did want to peer, and in some cases creative methods
were found to get something that looked an awful lot like "both peering and
transit", but for the most part abovenet was always seen by its customers
as an *alternative* to having to build a wide area network and employ BGP
engineers, since there would be just as much path splay at probably less
total cost and without the hassle of directly employing anybody who has
ever posted to NANOG.  (for the most part we don't dress nicely and are a
surely a surly lot, but don't call me shirley.)
-- 
Paul Vixie