UUNET Pulling Peering Agreements

Sean Donelan SEAN at SDG.DRA.COM
Sun May 4 02:40:33 UTC 1997


>One first thinks that it is only the little guys get burned with their
>customers when the little guy is seperated from a backbone operation such
>as UUNet's.  But the only thing is that out of the vast numbers of UUNet
>customers or indirect UUNet subscribers, there will be many who notice that
>a connection to the little guy is no longer possible -- and they will complain
>up the chain to UUNet.  Just responding to those complaints is a massive
>amount of work and expense.

Its only a massive amount of work and expense if you care about customer
service.  Remember what happened when users on PacBell Internet couldn't
reach the US Securities and Exchange Commission one morning.  The provider
that cared about its customers rectified the problem, at their own expense.
The provider that caused the problem, didn't suffer any reprecussions.  In
fact, they said "see it was the other guy's fault all along."

Stuff won't break all at once, it will just become more brittle.  It
will break weeks and months later, long after everyone has forgotten
that it could have stayed working.  I saw our first MCI hiccup a couple
of weeks after they disconnected from the CIX.  A different provider
turned down a BGP session at MAE-East, and has been straightening things
out for a week.  It is very difficult for a customer to isolate a problem
with the Internet, even when all the providers are cooperating.  Imagine
how tough its going to be for customers to figure things out with this
going on.

Providers that care about customer service are at a distinct disadvantage
playing chicken.  Providers that say "Damn the torpedoes, full speed ahead"
can get away with it.

I don't think anyone is looking for a "free" ride.  It would be great if
other providers compensated DRA for using DRA Net to carry their "free"
traffic.  But so far, none have made me an offer.
-- 
Sean Donelan, Data Research Associates, Inc, St. Louis, MO
  Affiliation given for identification not representation






More information about the NANOG mailing list