RE: PI/metro/geo [Re: The state of IPv6 multihoming development]

["Tony Hain" <alh-ietf@tndh.net>]

Michel Py wrote:
> ...
> I'm not sure I understand your question correctly. Slamming 
> as I understand it is make the customer switch to another 
> (yours!) long distance company in order to get the money.
> 
> We are talking about multihomed sites here, so a provider 
> that wants to slam would either:
> 
> a) Try to reduce the traffic on their customer's link by 
> providing routes that have crappy metrics (if the customers 
> pays a flat fee).
> 
> b) Try to increase the traffic on their customer's link by 
> providing routes that have very good metrics, which will 
> eventually push the customer towards a bigger, more expensive pipe.
> 
> Both of these are practiced today with IPv4. I'm not saying 
> it's a good practice, and one better have a good explanation 
> if caught, but it does happen. Are we looking at a solution 
> that *also* solves this problem?

I don't expect to solve it, but if we are talking about a system to
distribute mapping tables to align with current topology at the ingress
and restore the original values at the egress, we should make sure the
process does not make it easy to quietly influence which provider
carries the traffic. The obvious method to prevent this would be for the
site being mapped to sign its preferences, but that brings along its own
operational and scaling issues. 

Tony