Peering versus Transit

Mon Sep 30 05:28:29 UTC 1996

>Is it really the case that people with routers at exchange points actually
>consider a packet addressed to one of their own customers to be theft of
>service?  So far, I note, we haven't heard any position expressed by
>any of the big folks, just by others outraged on their behalf.

No, this is an attack.

If a packet goes where it is not supposed to (i.e. no route was explicitly
advertised) then the party who's sending that packet made an illegitimate
route.

That qualifies as a deliberate routing attack.  That is enough to get
kicked off any IXP.

Pointing default to somebody over IXP is simply theft.  In effect you
get somebody's router to sort your packets out.  That is covered by
applicable criminal statutes.  Needless to say, this is also a ground
for being kicked off IXPs.

Finally, most IXP operators do not seem to be diligent in explaining
to new IXP members to be that being at IXP is by no means a guarantee
that anybody would want to peer.

>And of course, to be in a position to "dump data" on a router at an
>exchange, one must have one's own router there peering with *somebody*,
>right?  So the problem, if there is one, is that not all pairs of
>routers at an exchange are necessarily allowed to communicate.  ATM,
>anyone?  (Gasp cough choke)

It is not a problem.  The problem is that people seem to have a hard time
understanding that Internet is not a free for all.  If two parties cannot
agree on terms of bilaterial exchange, that's their problem, ok?

There are two reasons for restrictions on who peers with whom -- first is
technical, as peering at exchange point is essentially excercise in trust;
and if somebody doesn't look like (doesn't have track record, have a person
with apparent drug problems as technical spokesman, etc) that they can
give you sane routes, and won't mangle routes you give them then big ISPs
are not willing to bet their business on that somebody staying sane.

The second reason is economical.  When players in one weight category are
talking to each other at many exchange points they don't subsidise each
other.  When a large provider talks to a regional provider at one IXP
it effectively spends much larger share of backbone resources delivering
traffic (for example, A is a member of Big 6, say B.F.C.; B is an ISP in
Mukhosransk, where one IXP is located.  When B's customer sends a packet
to A's customer in Lost Wages B pays for delivery from one corner of
Mukhosransk to another; A pays for getting it from Mukhosransk to Lost
Wages; it is exactly the same when packet is sent back.  Thus A pays
for disproportionate share of the communication cost.)

Obviously A does not have any interest in subsidising a competitor.

On the other hand, if B has particularly interesting content providers
which are in high demand by A's customers (say, Mukhosranskie Novosti)
A may decide that it is still worth to peer with B.

When there's C, of about the same size as A peering is obviously a matter
of mutual benefit; so B.F.C. and, say, A.F.T. won't have any problem
reaching a peering agreement.

So, this is very simple -- don't mess with the market.  It is healthy
as is.  Obviously big players want to keep their market share, but they
are not in collusion to keep others shut off.  A small provider can
choose whose customer he wants to be, and when he's grown up enough to
have population of his customer and content providers to be sufficient
leverage he will be automatically accepted into the "Big something".

A question with Europe is more interesting.  It is generally recognized
that Europeans pay more than their share when they drag bits to Boone Blvd;
so there should be no problem with peering save for bureaucratic
boneheadedness so cherished by some well-known companies.

--vadim