Peering Policies

Sean Donelan SEAN at SDG.DRA.COM
Fri Apr 23 04:55:30 UTC 1999

Why do these things always seem to happen in the spring time?

tim at clipper.NET (Tim Wolfe) writes:
>running over TCP seem to timeout with 2 second delays...)  So in order to
>maintain connectivity to both of these providers, one would need to buy
>transit from each.  How does this figure into the future political and
>technical growth of the net?  Wouldn't this type of apathy tend to push
>toward government involvement in this industry?  If the largest networks

I don't see why.  BT/Tymnet and GTE/Sprint/Telenet X.25 networks had
miserable interconnectivity, and many people ended up buying connections
to both.  Or using the Internet instead, which killed both the x.25
network providers.

Regulators did step into the telegraph business and require Western Union
interconnect with its competitors, just about the time telegraphs ceased
as a viable business.  ILECs and CLECs are required to interconnect, but
that doesn't seem to end problems with lack of inter-office trunks, restrooms,
parking lots and other shenanigans.  Government involvement doesn't seem
to solve the problem.  Settlements don't help either, because as we've
seen in the ILEC/CLEC cases, one side just simply refuses to pay if it
doesn't go the way they wanted.

What has worked?

Bad press, and customer reaction have stomped most of the peering battles
in the past.  After a big storm, both sides will issue non-specific
press releases saying everything is hunky-dory again.  As engineers, we
might think there must be a better way.   Sometimes the problems aren't
technical, and don't have an engineering solution. Master tickets will 
get escalated to the seventh level senior network gods, and sit indefinitely.

Ebbers and Esrey may not know a packet, but when the CEO's get asked about
it on CNNfn, it will get fixed darn quick.  If it doesn't get fixed?  Well,
does anyone even care what AGIS's peering policy is anymore?

amb at gxn.NET (Alex Bligh) writes:
>If you order the 5 largest IP networks by traffic profile, you
>may get some surprises.

Porn and Spam for $500!

Traffic profiling will show there is a difference between where people
say they go, and where they actually go.  And the top 5 networks are not
the ones you would expect.  Of course, NDA's cover most of that information.

>From a traffic engineering standpoint, the Internet is still very much
a network of networks.  I wouldn't sign any long term contracts with one
provider just yet.  Just look back 1, 3, 5 years at the changes in who's
Sean Donelan, Data Research Associates, Inc, St. Louis, MO
  Affiliation given for identification not representation

More information about the NANOG mailing list