[ppml] Policy to help the little guys

[David Williamson]

On Wed, Mar 19, 2008 at 04:17:32PM -0400, William Herrin wrote:
> The difference between PA and PI is an artificial result of the RIRs'
> upside down solution to the BGP cost problem. You can't get PI because
> you probably won't put enough money into the backbone infrastructure
> to fairly cover the cost of your being there while someone getting PA
> probably will.

I'm entirely baffled by this assertion.  My site uses PI space, but
we're highly multihomed and default-free.  We clearly pay the same
costs to participate in the DFZ as anyone else.  Our upstream peers
charge us for bandwidth and the privilege of peering with them.  We
have routers of sufficient scale to support our existence in the DFZ.

Your assertion seems to imply that any org like mine isn't paying it's
fair share, but we would be with the exact same setup in PA space.  In
other words, if we got space from one of our upstreams, we'd magically
be paying our share.

I don't doubt your analysis shows that inserting a route in the DFZ
causes a bit of expense for everyone else in the DFZ, but I don't see
how PA versus PI has anything to do with it.

The only way I see PI as destructive in the way you describe is for a
PI site to be singlehomed with a default route to their upstream and an
announcement of their AS and IP space.  In that case, you need a Cisco
2514, and you are probably not bearing your fair share of the costs of
your route.  I've already stated that I don't understand the utility of
singlehomed PI.  In the multihomed case, however, I think your assertion
in the quoted message is entirely incorrect.

I've also hit my own quota for posts in one day, so I think I'll let
others chat on these topics for a bit.

-David