Intelligent aggregation techniques break MEDs and therefore introduce a need for advertisement of more-specifics in order to convey the optimal entry point into an AS. Then... Poor address allocation/aggregation models and broken IGP metric -> BGP MED associations, coupled with fear that the peer network will "leak the more specifics", make questionable the value of MEDs, especially for "well established TIER 1" providers with tons of legacy stuff or a really large customer base. -danny
or why dont you require both to peer in 4 - 6 places in the US and honor meds both ways. This way, each is carrying traffic on their network. Of course the bigger providers dont want this as they would lose losts of $$ from all the payments they get for 'transit'.