Re: comments on draft-py-multi6-gapi-00.txt

[Brian E Carpenter <brian@hursley.ibm.com>]

Iljitsch van Beijnum wrote:
> 
> On Thu, 6 Feb 2003, Joe Abley wrote:
> 
> > Interconnecting countries in South Asia and other developing regions is
> > most definitely a governmental policy issue, and not a technical issue
> > (the technical case for direct interconnection is trivial, and yet
> > there is no direct interconnection). It might take governments twenty
> > years to decide that fibre across the land border is a good idea, or
> > they might not decide to do it at all.
> 
> Both goverments and business make bad decisions at times. Only
> governments usually need more time to make them.
> 
> However, there may not be a technical case for interconnection. Here in
> the Netherlands around 45% of all traffic is to the US (this includes
> stuff that is only reachable through the US), 45% is domestic and 10% to
> the rest of Europe. Obviously these aren't very hard figures, but each
> time I checked it was pretty close. Under these circumstances it doesn't
> make much sense for non-huge networks to have direct connections to
> neighboring countries: reserving some extra capacity on the US link
> makes much more sense as it is virtually free, no extra effort and
> better burstability. Now recompute with much higher costs for links to
> neighboring countries and lots of government intervention.
> 
> > Any routing system that relies on (geopolitically) local
> > interconnection better be able to accommodate frequent departures from
> > that model. There are a lot of people in South Asia, and the
> > infrastructure is growing a lot faster than governments are able to
> > think, never mind legislate.
> 
> In the US there isn't much local interconnection as not much traffic
> stays local. In most other countries, much more traffic stays local
> because the countries are usually simply smaller and because of language
> barriers. So now we have examples with little local interconnection and
> examples with little regional interconnection. I'm still waiting for
> examples of significant internet use with no local OR regional
> interconnection.
> 
> Not that any of this is of practical importance with regard to the
> decisions we are facing.

On the contrary, it all illustrates that there are very limited
economic forces driving the topology towards geographically-based
interconnects. In fact, we basically don't understand what caused
the various kinks in the bgp.potaroo.net curve, except for the one
at 20k announcements (i.e. the BGP4 switch on).

I can't disagree with the assertion that *if* we start handing
out PI prefixes, a loosely geographical or metro-based solution
is probably better than random numbers. But until we actually
know how PI prefixes are going to become routeable without
creating a swamp, I'd rather sit on my hands.

   Brian