Re: Regionally aggregatable address space for multihoming

[Geoff Huston <gih@telstra.net>]

> > how do you communication that information to the remote system in a way
> > that manages to constrain the choices available to the remote system to
> > match your desired policy?
>
>when 'remote' means they have no technical or financial incentive to do you
>the favor, then 'constrain' seems far too strong a term.

Indeed when favours run out, then constraining choices is precisely what is 
going on. i.e. there are no technical or financial incentives for the party 
who is sending the traffic to take the destination's preferred path, but 
there are financial incentives for the destination to coerce the sender to 
make a particular choice. Interdomain traffic policy always struck me as a 
variant of the Dyjkstra (sp?) brides problem - the starting case is that 
the sender has unilateral choice as path choice to the destination, and the 
destination responds to this by attempting to constrain the sender's choice 
set. This game is being played out in the inter-AS space because, like the 
commons, there is no impediment to abusing this common space.


> > BUT you also have to work out how to overlay policy / TE onto of this
> > connectivity fabric
>
>i have an arch-capitalist friend who i would normally expect to step in
>here and observe that it is not clear that you HAVE TO.  in fact, it might
>be in the interest of only a few.  perhaps, when it comes to traveling this
>road, he has fallen arches, or maybe it is feet of clay, or fear of capital
>punishment. :-)

If you want the interdomain connectivity protocols to scale then separating 
out policy / TE 'negotiation' into a distinct protocol interaction would 
address some of the concerns regarding scaling of the inter-domain space as 
a connectivity protocol convergence issue. The imperative of "HAVE" comes 
for broader issues surrounding the integrity of the whole, as represented 
the sum of all these individual policy-based transactions.

As an addendum to this posting I would have to say that I think I'm abusing 
the tolerance of the multi6 list with this topic, and I promise that I'll 
stop. Please accept my apologies if you already think I've gone well out 
into the weeds topic-wise. As interesting as these topics are to me, I 
think that I've taken up more than enough of the multi6 crew's time, and I 
try and stay focused on multi6 in further postings to this list.

Randy, Do you have any suggestions as to an appropriate list to continue 
this discussion? Is ptomaine willing to tolerate a continuation of this 
discussion?


thanks,

   Geoff