She can't give the school permission to access Facebook's computers, only Facebook can do that.

Of course not. Paying for bandwidth asymmetry is only used where its logic makes sense. The basic underlying assumption is that traffic that begins on one network and terminates on the other benefits both sides equally and thus the costs should be split roughly evenly. Paying for bandwidth asymmetry is an approximation to cover the case where one side pays more than the other, usually because one side has to carry the traffic further than the other. (Generally, you carry inbound traffic further than outbound.) Historically, this is the way it's done.

But when you're talking about Comcast, which has a large number of small endpoints, and Netflix or Crashplan, which have a small number of large endpoints, more of the costs are borne by Comcast regardless of the direction. So settlement-based peering makes sense regardless of traffic ratios.

Settlement-based peering based on traffic ratios makes sense when you're talking about two ISPs with roughly similar business models, types of customers, and service areas. But it's just a simple approximation of the underlying logic -- traffic benefits the sender and receiver about equally, so they should split the costs about equally. When design asymmetries make one party pay more than their fair share, settlement-based peering is the norm.

Comcast is not an end user, they are a peer. When two networks exchange traffic as peers, that means they exchange only traffic that originates on one of their networks and terminates on the other. This is precisely what Comcast and Netflix want to do -- exchange traffic that originates on one network and terminates on the other. That is, by definition, peering.

I'll put this in terms that are as simple as possible, since you clearly don't understand the difference between peering and transit. When you want to reach a network in Sweden, your ISP carries that traffic for you. When your ISP wants to reach a network in Sweden, they can't ask you to carry it. That's why you pay your ISP. It has nothing to do with ratios or directions -- it's because your ISP is providing you with transit and you are not providing your ISP transit.

The presumption behind the scheme is that the traffic benefits both sides equally and thus the costs should be split. If Netflix wants to receive traffic that provides no benefit whatsoever to Comcast customers, they should pay 100% of the costs for that traffic. So that won't actually even the flow at all but only make it more asymmetric.

You don't have a peering arrangement with your ISP, you have a transit arrangement with them. These things are *completely* different. You use your ISP to reach other networks. Netflix doesn't want to use Comcast to reach anyone else.

I think you don't understand what *peering* is.

They're paying twice, but they're paying for two different things. Netflix pays their ISP for the cost of delivering their traffic to Comcast. But they also pay Comcast for half the difference between their ISP's delivery costs and Comcast's delivery cost.

Say I want to send a package to you and we agree to split the cost of delivery. I might pay a courier to take the package to a pickup point and you might pay a courier to take the package from the pickup point to you. But if your courier is more expensive than mine, then I should also pay you half the difference in costs, so that we fairly split the delivery cost. This is the basis of settlement-based peering.

No, you should pay half the cost of the delivery. The sender should pay the other half. This is why the Internet has, for decades, used settlement-based for asymmetric flows. Otherwise, folks like Netflix and Google won't be paying half the costs.

It is fundamental to Netflix's business model that they will have a small number of sources that produce large amounts of traffic to a large number of destinations. It is common sense that this means that Netflix will, in the absence of settlements, pay less than half the costs of the traffic they produce.

"The ISPs aren't peering at all, they are the termination point."

Huh? Peering is how traffic is exchanged between termination points. Traffic from Netflix to Comcast customers can reach Comcast's network three ways:

1) Netflix can peer with Comcast.

2) Netflix can be a transit customer of a network that peers with Comcast.

3) Netflix can be a transit customer of Comcast.

1 is what they agreed to now. 2 is what they had before using Cogent's peering with Comcast. 3 is unreasonable given the business nature of Comcast and Netflix.

> Why should Netflix hesitate to explain to their subscribers that their ISP is the cause of poor streaming because they refuse to peer with the Netflix network?

They were happy to peer with Netflix, Netflix just didn't want to pay for it. They tried to get Cogent to bully Comcast and that failed. So Netflix agreed to peer with Comcast directly. Settlement-based peering for traffic imbalances has been the norm for more than a decade. Netflix and Cogent just tried to strongarm Comcast and failed.

Google and Facebook produce large amounts of outbound bandwidth from a small number of endpoints. Comcast handles large amounts of inbound traffic to a large number of endpoints. The traffic benefits both parties equally, but it is inherent in both Google's and Facebook's business models that their traffic will cost them less than the costs it imposes on the destination networks. Thus it makes sense for Comcast to charge them to balance things out. This argument doesn't work in reverse.

You're tilting at windmills here. It has been a very long time since "Internet access" meant unfiltered, unthrottled access. If you were arguing for laws requiring clearer disclosure about exactly what people are getting, I wouldn't have a problem.

I have no problem with a law requiring carriers who claim to be neutral to in fact be neutral. But otherwise, this is like the nutty argument that fractional reserve banking is inherently fraudulent because some people might think the bank keeps all the deposits in their safe.

Right, but what about when there's *ZERO* choices for Internet? Presumably, anti-net neutrality policies are profitable for ISPs or they wouldn't want to adopt them. So you've just discouraged the very competition you claim you want by reducing the profitability of entering underserved markets.