Yeah I stopped reading at that point. An ISP with a history of interfering with certain traffic and lobbying against net neutrality is not a supporter of an open Internet at all, let alone the "strongest."

Myth: "No ISP in the country has been a stronger supporter of the Open Internet than Comcast[.]"

Fact: "Comcast spent the most money of any organization in support of the Stop Online Piracy and PROTECT IP bills, spending roughly $5 million to lobby for their passage."

http://en.wikipedia.org/wiki/Comcast#Lobbying_and_electoral_...

That's what's really bothered me about this whole thing--Comcast and the likes have already mastered their phrasing, and that bodes poorly for public perception of the issue. I've heard the term Open Internet used in that context several times in the last few days, and I think it's a part of a concerted effort to paint preferential traffic treatment as making the internet more "open for business". The hope is that rosy phrasing will "help" Joe Shmoe decide that the FCC's proposal is pro-consumer and pro-openness. They're not using the term in some cynical attempt to say that it ain't so, they're trying to actively re-define and capture it to form a cohesive narrative.

I was just about to post that exact same comment. I nearly spit out my drink reading that line!

Agreed, I just came to comment on that sentence as well. It's presented with no argument or references so it's just hyperbole.

And is Comcast supposed to wire up every server for free? If I stick up a box and slap on a 1Gbps uplink to some hosting company that only bought 100Mbps from Comcast, should Comcast now give my and my hoster free access because you decided to download a video file from my server?

That's not what's expected. What's expected is that you get what you pay for, whether you're a customer on the hosting end or the residential end. If you're paying your data center for 1Gbps, then they ought to have at least 1Gbps capacity to the major residential ISPs. If you're paying your residential ISP for 15Mbps, then they ought to have at least 15Mbps capacity to their peers. If there are 100 people trying to use 15Mbps at once at peak hours, they need at least 1500Mbps of capacity to provide the level of service they sold.

What Comcast is being chastised for, the thing that forced Netflix to transit with them directly, is being paid by its customers to connect to the internet, then refusing to use that money to buy a pipe large enough to deliver that. It wasn't Netflix that wasn't buying a large enough pipe to send its own bits. If Comcast had resolved its disputes with Cogent and widened the pipe to meet the customer demand it's being paid by those customers to meet, I'm sure Netflix would've been happy to stick with its ISP.

tldr: peering issue, right?

google has similar issues with some EU providers in particular in france, for youtube.

traditionally the ISP pays to upgrade the pipes for the content that is provided by a third party (ie ISP pays for youtube, netflix traffic). ie the one whos downloading pays for upgrading pipes.

thats where some ISPs (comcast too then i guess?) generally cringe: this way of dividing costs was ok 10 years ago, but not anymore: streaming is 90% of their traffic and its not sustainable.

I think Comcast, owner of NBC Universal, has little ground to stand on here. Streaming is an inefficient use of the Internet that is popular only because people are too terrified of lawsuits to use better technologies like BitTorrent. Comcast has worked hard to set up the legal environment that created this situation; they will see no sympathy from me.

In most cases, streaming is basically no different than downloading, right? The vast majority of people watch a video only once.

So it's more like saying "downloading is 90% of their traffic and its not sustainable."

Sure, but it doesn't have to be that way. Something built around multicast groups could be used to stream multiple people the same content in a drastically more efficient manner, and then it could be stored locally on clients for time-shifting purposes.

If you're trying to optimize for efficient use of limited bandwidth, unicast transferring of identical content to many many people is pretty wasteful. I think Netflix would argue that network links should be getting bigger and fatter to render the point moot, but given the streams are also getting bigger (we didn't always stream 720p everywhere, did we?) that would certainly take a lot more investment than what's happening now.

That's not true. Backbone pays for long haul to meet me room. ISP pays for last mile from meet me to consumer. Comcast decided to not only not pay for interchange equipment, but also to charge twice for traffic. Classic monopolist move.

It's worse than that. If you look at how much bandwidth is required for any-to-any connectivity you can approximate how much it costs by modelling the internet as a full graph. Of course the cost of a full graph internet is n(n-1), with n the number of customers.

The problem is, what happens when you connect a new customer. Well, simple :

1) the income of the ISP increases by some fixed amount

2) the cost of providing this connection increases by n+1 (the number of existing customers)

So here's the catch

np needs to pay for n(n-1)c pipes

Where n is the number of customers on the internet. P is the average income per customer and c is the cost of a pipe that will sustain one customer.

Needless to say, this won't work.

Approximating the Internet to a complete graph with edges of equal cost is absurd, that's not how it operates at all.

True, but people claiming: "I pay for XXX Mbit/s, I want my XXX Mbit/s 100% of time to 100% of the Internet" think/want the Internet to operate like that.

> If there are 100 people trying to use 15Mbps at once at peak hours, they need at least 1500Mbps of capacity to provide the level of service they sold.

No retail ISP on this planet offers 1:1 ratio like that to its general subscribers.

You can certainly buy uncontended service, but it will start at several hundred dollars per month for about 2 Mbps.

Unfortunately you don't buy an always open 15 Mbps pipe, you buy "15 Mbps peak speeds", or at least that's how it is worded for me. Which means at peak usage times you expect to get lower than 15 Mbps.

It's hard for customers to buy that when the YouTube and Netflix videos they're watching only need a few hundred Kbps, but they can't get one good stream going on an otherwise unutilized connection. It's not a capacity crunch near the edges of Comcast's network, where lines are still shared in neighborhoods and upgrades cost the most; it's just greed, fattening already healthy profit margins by refusing to increase peering capacity and demanding peers pay more for bits customers are already paying for.

It is worse than that. "15 Mbps peak speeds" is actually "15 Mbps bursts."

I am fine with getting less than my maximum speed; I understand that the connection is shared and that my neighbors' service can interfere with mine. TCP can deal with congestion. What grinds my gears is that ISPs are not content to sell me a maximum speed; they want to oversubscribe, then claim that they are selling burst service, then call it "unlimited super-awesome Internet!! 15Mbps!!" then penalize customers who fail to limit their use of the supposedly unlimited service.

Which means deciding what an acceptable QoS is for those peak times - many of us think they decided too low - I personally think you should be able to get between 1/3 and 2/3rds (ideally around half) of 'peak' speeds on a sustained level at any time for any amount of time.

It's just like the telephone network of yore, no single trunk was ever capacitied to mothers day call traffic - there were reroutes, and more than one path to any large point, so you'd get a lower quality of service, but the traffic would still get thru.

> should Comcast now give my and my hoster free access

If your hosting company has paid Comcast for 100Mbps transit, then Comcast should honor that agreement with your hosting company. Your uplink to your hosting provider is completely irrelevant as far as Comcast is concerned.

If you're using a different transit provider (e.g., Cogent), then Cogent and Comcast will have a mutual peering agreement that accounts for imbalances in transfer. These details are very complicated, proprietary, and you should never need to care about them. That's the whole point of the internet; you connect to "the internet" and can treat it as a homogeneous network, whereas in reality it is actually a bunch of mutually-peered subnetworks.

>Cogent and Comcast will have a mutual peering agreement

Yeah that's what all these disputes are coming down to. I can brand myself an ISP and sell someone a 10G uplink. That doesn't magically make anyone have to peer with me. Yes it's complicated and yes it comes down to who can throw weight around.

In your example the hosting company is also a Comcast customer, so that hosting company needs to upgrade its link. That isn't the case here. Netflix pays for its own network infrastructure to get the content to a neutral exchange point and is willing to upgrade that infrastructure as needed. Comcast should be willing to upgrade the infrastructure from that point to its customers houses as that's what their customers pay them to do.

If the ISP was a customer buying bandwidth from Comcast like anyone else, then Comcast has zero obligation to increase their bandwidth for free.

If the ISP is Comcast, then when you pay for that 1Gbps uplink, you are paying Comcast and you have a right to expect that your packets will be delivered at that rate.

If the ISP is a peer of Comcast, then it seems like when you pay that ISP, you ought to know the exact details of the agreements and bandwidth arrangements between the ISP and Comcast, and both parties ought to abide by that contract. I don't know enough details in this case to know whether that's what happened.

It's even better/worse than that. If Comcast customers are requesting 1Gbps of traffic from your server, Comcast may need to buy transit to ensure those packets aren't dropped.