This thing between Level 3 and Cogent isn’t the first peering issue. There have been others, like when several backbone carriers threatened to start charging to transit non-local traffic. I think that was MCI, which at the time essentially was the backbone. Slightly different was the issue when some podunk ISP in Florida had munged BGP locally and the damage got propagated such that huge swaths of the Internet got routed down to some tiny ISP …
This is what the public peering points were supposed to be for, but they were always seriously overloaded. Internap’s big entry into the market was all about private peering agreements in order to bypass the public peering points, in order to route more efficiently.
But, here’s the delayed consequences of relaxing the commercial rules about the use of the Internet back before the big expansion in 1995. I’m not saying the change was necessarily bad, but the public and open infrastructure has become increasingly privatized and negotiated. If two companies can de-peer each other and isolate entire islands of the Internet, then the power of the protocol to route around fault is at risk. The inherent strength of the Internet to withstand disruption of specific nodes and routes is compromised. If the physical structure of peering is such that it is no longer possible to route around damage, then the whole infrastructure is at risk.
Google has been buying dark fibre, and there’s a whole lot of dark fibre. Fibre in the ground was one of the big projects during the technical boom, but it’s managed like the diamond or oil supply: supply is kept back in order to manage prices and profit. So lots of fibre was put in the ground, but left dark. Now, Google has been buying it. They’ve certainly got reason, by way of massive amounts of content being served, to want their own network with private peering agreements.
However, it seems to me that dark fibre should really become the new highway system, not even more private toll-roads on an already endangered system of information transportation.