How could your ISP possibly be a third party network? You're a directly attached user on the ISP network, within their management and addressing domain. The network exists explicitly to move your bits, and you're paying for the service. A third party network is one moving bits between stations that aren't part of that network.

If you ran a big network that Google needed to transit through to get to stations beyond your network, then yes, there'd be a transit settlement in place. That's how the Internet works. If you ran a big network, and your direct users requested access to Google resources, then no, Google would not pay you to deliver the traffic that your users request. You're soliciting the traffic on behalf of your customers, who pay you to do so. There's no reason why Google should pay you their money, and use their resources to fulfill a request that your users are paying you to facilitate.

If you really were to argue that Google should pay, then where do you mark the line? How much traffic warrants payment? Who has to pay? How big do you have to be? The entire Internet would become analogous to people subscribing for PSTN service and then placing collect calls to everyone they want to talk to.

That's a benefit of being a large content provider. You become the tier 1 networks' product. Just like you become the end user ISPs' product. Notice, however, the distinction here. Regardless of whether or not a small company has to pay for their transit, it certainly doesn't have to pay for access to Orange's end-user network. Nor should Google, or any other company whose resources are being requested by the users of Orange's network.

>Which one does Google sending data to me via my ISP come under?

Where is the source of confusion? Google is a customer to its transit providers. To your ISP, they're just another station on the Internet sending bits that its clients requested.

>Seems a bit of an arbitrary distinction, unless we want to go by technicalities like "having your own IP range / AS number makes you a separate network".

There's nothing arbitrary about it. If your traffic needs to traverse a third party network, then you pay that third party network for the privilege. It's very straight-forward. There is absolutely no source for confusion in any of this.

>The "users" want to access resources. Google want them to, so they can sell advertising. The benefits of the transaction aren't one-way

Nor is selling milk. The customer gets milk, the dairy gets money.

>Besides, if I used my internet connection exclusively to host a server I would still have to pay for it. The difference is that Google is big and important enough to be able to bargain for good terms.

You pay for transit as a provider. Transit is when your traffic passes a third party network. Transit is not when the traffic that the users of a network requested passes through their own network.

At the request of their users. Just like every other part of the Internet works. ISPs make money because their users want to access resources, including Google's. Charging Google for the traffic generated by the requests sent by the subscribers who are also being charged is like the grocery store charging the dairy farm for the shelf space to stock their milk.

XKCD is a comic that involves a lot of math and science, and complains about dishonesty in academia on a frequent basis. Yet here the author is being excessively generous to one concept that he favours, and excessively pessimistic regarding another that he's trying to dispel. That's what irks me.

But if you don't understand the concept, and would rather write a post full of meaningless insults and devoid of reason and sensibility, then there's nothing anyone can do to stop you. Knock yourself out.

I read the transcript. "A few more bits" would actually be a good few more bits. There are many, many more feasible and realistically common permutations. Enough, in my opinion, to bring the total entropy to a level way beyond what's necessary to use on any system that allows a thousand queries per second.

That's a rather offensive opening for a post that doesn't really do anything to corroborate the methods used to calculate the entropy. I'm well aware of the exponential benefits of longer passwords.

That's the real reason to pick a long password, yes, but I have issues with the way the entropy is calculated in that strip. It's based on assumptions, and the entropy seems to be calculated with no regard to any of the many other common permutations.

I seem to be the only one who thinks that this strip is complete bullshit. There are so many assumptions, and no accounting for variations.

Don't forget uc "free";

I'm having trouble understanding your reply. The people who play games don't buy $300 video cards to run Photoshop. They buy them to run very demanding 3D applications. You can't run those applications on other platforms, so it isn't a matter of going out of one's way to avoid alternatives.

"High performance desktop" gamers are a pretty self-selected group now days. If you're sitting in front of a ATX case with a discreet video card, you've gone out your way to avoid every computing trend over the last 10 years. Which is fine, but its not exactly a growth market.

That's a pretty loaded statement. If you use a tool for a specific tasks, and forgo newer tools that come out in favour of revisions of the tool that you have been using because it remains the best tool for the job, then you haven't "gone out of your way to avoid every computing trend," rather, you've continued to use the best tool for the job. There are no devices more suitable for the kind of stuff these people do than desktop computers with discrete video cards.

Somehow this still costs more than what we pay T-Mobile for two phones with unlimited minutes and text, and 2GiB for each phone.

The U.S. is a third world country with respect to cellular providers.