Maybe it would help if you tried to address the issues in a serious way instead of just trying to be cute.
Just a thought...
RB
I don't think it's conflation, Joly, since the essence of NN is for the eyeballs to pay for the entire cost of the network and for edge providers to use it for free; isn't that what Netflix is asking the FCC to impose under the guise of "strong net neutrality?" Professor van Schewick is pretty clear about making the users pay for the edge providers in her tome on Internet architecture and innovation.
Competition is a wonderful thing where it can work, but it's not a panacea, especially for the poor and for high-cost, rural areas. Communication policy has pretty much always relied on some form of subsidy for these situations, that's the universal service fee we pay on our phone bills.
Susan Crawford explicitly complains that American ISPs "gouge the rich" by charging more than the OECD norm for high-speed (50 Mbps and above) service, but she fails to point out that they also charge less than the norm for low-speed (15 Mbps and below) service.
I think it's easy to create unintended consequences if you don't look at how specific regulations affect real people, no matter how high-minded and principled they may appear at the surface.
RB
I won’t presume to speak for Netflix, and I won’t presume to provide a canonical definition of “network neutrality.” However, I can say what global prevailing business practice is, since I’ve actually surveyed and quantified it:
Each network (regardless of whether they term themselves “eyeball,” “content,” “edge,” or whatever) delivering a packet pays their own way to the IXP of their choice that the other party is present at, each network receiving a packet pays their own way from the IXP of their counterpart’s choice that they’re present at, independently in each direction. Thus, where content networks interconnect with eyeball networks, when they follow the best practice engaged in by 99.73% of all network-pairs, the eyeball network’s customers pay them to deliver traffic to an IXP of their choice and from an IXP of the content network’s choice, while the content network’s customers pay them to deliver traffic to an IXP of their choice and from an IXP of the eyeball network's choice, long in, short out. No money changes hands between the two networks, because no value is exchanged between the two networks. Each network pays their own way, and is in turn paid by their customer. Because they’re each providing value to their customers, not to each other.
In 0.27% of cases, the parties aren’t able to see their way to following best practices, and some fraction of those are disputes between content and eyeball networks of the sort that you’re describing.
-Bill
Bill Woodcock wrote:
The essence of NN is for the eyeballs to pay for the entire cost of the network and for edge providers to use it for free; isn't that what Netflix is asking the FCC to impose?
I won’t presume to speak for Netflix, and I won’t presume to provide a canonical definition of “network neutrality.” However, I can say what global prevailing business practice is, since I’ve actually surveyed and quantified it:
Each network (regardless of whether they term themselves “eyeball,” “content,” “edge,” or whatever) delivering a packet pays their own way to the IXP of their choice that the other party is present at, each network receiving a packet pays their own way from the IXP of their counterpart’s choice that they’re present at, independently in each direction. Thus, where content networks interconnect with eyeball networks, when they follow the best practice engaged in by 99.73% of all network-pairs, the eyeball network’s customers pay them to deliver traffic to an IXP of their choice and from an IXP of the content network’s choice, while the content network’s customers pay them to deliver traffic to an IXP of their choice and from an IXP of the eyeball network's choice, long in, short out. No money changes hands between the two networks, because no value is exchanged between the two networks. Each network pays their own way, and is in turn paid by their customer. Because they’re each providing value to their customers, not to each other.
In 0.27% of cases, the parties aren’t able to see their way to following best practices, and some fraction of those are disputes between content and eyeball networks of the sort that you’re describing.
-Bill
Bill,
Can you say more about what you've done to "survey and quantify" prevailing practices?
And... given that Netflix is reportedly about 1/3 of Internet traffic these days, and Verizon is huge - how does that come out to .27% of cases (leaving aside other recent disputes like L3-Cogent, and Netflix-Comcast)?
Miles Fidelman
I will when you will, "poopy head".
- Matt
In a word: no. Net neutrality is about everyone paying their own way to get
their packets to where they want them to go. Netflix doesn't get to use the
Internet "for free"; they pay a whole heck of a lot each month to L3 and
Cogent.
- Matt
Can you say more about what you've done to "survey and quantify" prevailing practices?
https://www.pch.net/resources/papers//peering-survey/PCH-Peering-Survey-2011.pdf
We’ll do another one in the run-up to the next OECD carrier interconnection paper.
Given that Netflix is reportedly about 1/3 of Internet traffic these days, and Verizon is huge - how does that come out to .27% of cases?
Netflix/Verizon would be 0.0007% of cases, if it’s represented in the dataset. The survey was of interconnection norms, not of hugeness.
-Bill
In fact Netflix is asking to connect to eyeball networks for free:
http://blog.netflix.com/2014/03/internet-tolls-and-case-for-strong-net.html
" Strong net neutrality additionally prevents ISPs from charging a toll for interconnection to services like Netflix, YouTube, or Skype, or intermediaries such as Cogent, Akamai or Level 3, to deliver the services and data requested by ISP residential subscribers. Instead, they must provide sufficient access to their network without charge."
This isn't the traditional understanding of net neutrality, but this is the beauty of murky notions: they can be redefined as the fashions change: "You've designed your network to handle the traffic demands of web browsing? That's cute, now rebuild it to handle 40 times more traffic while I sit back and call you a crook for not anticipating my innovation."
Very wow.
RB
Now that's more than a little disingenuous. Until a week or so ago, pretty much all of the FIOS plans were asynchronous - a 15meg down/5meg up network was not designed for web browsing and email.
For that matter, Verizon is currently billing their lowest speed FIOS plan, at 50up/50down as "Stream 2 HD videos simultaneously" and for only $20/mo. more you can "stream up to 7 HD videos simultaneously"
Miles Fidelman
Richard Bennett wrote:
You are aware that there are, probably, thousands of eyeball networks
doing this right now, right?
Drive Slow,
Paul Wall
In fact Netflix is asking to connect to eyeball networks for free:
http://blog.netflix.com/2014/03/internet-tolls-and-case-for-strong-net.html
" Strong net neutrality additionally prevents ISPs from charging a
toll for interconnection to services like Netflix, YouTube, or
Skype, or intermediaries such as Cogent, Akamai or Level 3, to
deliver the services and data requested by ISP residential
subscribers. Instead, they must provide sufficient access to their
network without charge."
The important phrase there is "requested by ISP residential subscribers".
You will see this material again.
This isn't the traditional understanding of net neutrality, but this
is the beauty of murky notions: they can be redefined as the
fashions change: "You've designed your network to handle the traffic
demands of web browsing? That's cute, now rebuild it to handle 40
times more traffic while I sit back and call you a crook for not
anticipating my innovation."
A more accurate phrasing would be, "You've designed your network to handle
the traffic demands of web browsing, while *telling your customers they can
stream video*? That's cute, now provision a few more circuits to your
upstreams to handle the traffic that you said you could handle, instead of
trying to leverage your monopoly position to rent-seek off me."
Entrenched monopoly is what this is all about, ultimately. Nobody in
Australia (my home town) talks about Net Neutrality. We don't care. We
don't *have* to care. Because no ISP over here currently has a sufficiently
captive market to permit them to play chicken with a content provider. Any
ISP who did, and held their customer base to ransom, would very quickly find
themselves losing customers -- at least that segment of the market that used
the relevant content provider's services. Perhaps that wouldn't be a bad
thing for the ISP -- less traffic, lower costs, better margins... but at
least customers would be able to choose. No such luck in the US, where some
eye-wateringly high percentage of users have no choice in who provides them
a given service.
- Matt
Wait, I'm confused?
Of the ISPs can't handle 5mbps of traffic when a customer wants to watch
TV, why the hell are they selling 100mbps plans!?!
Answer that with something other than "because the ISPs more lucrative
content business is threatened by Netflix"?
Stop trying to hide what this so obviously is.
Others:
Do you know if Netflix peers with tier 1s (level 3, cogent, etc) or
purchases capacity?
Bennett:
Sorry for the double mail, still getting used to gmail on the Android.
Jed Robertson
route-views will confirm that Netflix peer with a number of access
providers, including the large ones; press releases related to
"OpenConnect" imply that no money is passing hands.
You'll note that, in spite of his wordy replies, never once does
Richard Bennett disclose who is funding him and AEI. Call it whatever
you want, I think "lobbyist" is the best word choice.
Drive Slow,
Paul Wall
Bill Woodcock wrote:
Can you say more about what you've done to "survey and quantify" prevailing practices?
https://www.pch.net/resources/papers//peering-survey/PCH-Peering-Survey-2011.pdf
We�ll do another one in the run-up to the next OECD carrier interconnection paper.
Interesting study. Thanks for the pointer.
Given that Netflix is reportedly about 1/3 of Internet traffic these days, and Verizon is huge - how does that come out to .27% of cases?
Netflix/Verizon would be 0.0007% of cases, if it�s represented in the dataset. The survey was of interconnection norms, not of hugeness.
It is worth noting, though, that not all interconnection are created equal. I wonder how your numbers would come out if you grouped interconnection agreements by amount of traffic exchanged, level of asymmetry, and so forth. And then perhaps by level of competition in the associated markets (do monopoly carriers behave differently than ones where there is a lot of competition?).
Just by analogy, the answer to "what kind of protocol traffic dominates the net" (or is "more important") differs considerably if you look at bandwidth vs. transactions (last time I looked, admittedly a little while ago, email still dominates network traffic when you look at transactions; but video clearly eats of most of the bandwidth).
Regards,
Miles Fidelman
Paul WALL wrote:
route-views will confirm that Netflix peer with a number of access
providers, including the large ones; press releases related to
"OpenConnect" imply that no money is passing hands.You'll note that, in spite of his wordy replies, never once does
Richard Bennett disclose who is funding him and AEI. Call it whatever
you want, I think "lobbyist" is the best word choice.
It's pretty well established that AEI is primarily a right-wing, conservative, pro-business think tank - with a mission statement that starts: "The American Enterprise Institute is a community of scholars and supporters committed to expanding liberty, increasing individual opportunity and strengthening free enterprise." (http://www.aei.org/about/)
AEI policy studies are pretty consistently anti-regulation.
Miles Fidelman
However, I can say what global prevailing business practice
is, since I’ve actually surveyed and quantified it:Each network [..] pays their own way to the IXP of their
choice that the other party is present at, each network
receiving a packet pays their own way from the IXP of
their counterpart’s choice that they’re present at,
independently in each direction.
Hi Bill,
I take issue with this claim because:
The survey was of interconnection norms, not of hugeness.
And, "Of the total analyzed agreements, [...] 141,512 (99.51%) were
“handshake” agreements in which the parties agreed to informal or
commonly understood terms without creating a written document."
As a result, the data set suffers three flaws:
1. It is not representative of the actual traffic flows on the Internet.
2. The overwhelming majority of the agreements analyzed were handshake
agreements but no picture is available of the handshake agreements
those same parties rejected outright or, expecting rejection elected
not to pursue. That creates a data bias which could mask any number of
factors, leaving you no way to determine that the claimed norm bears
any resemblance to the results one might expect when proposing peering
with a neighbor.
3. The data supports no affirmative statement about the the peering
case most relevant to network neutrality: that of a small network
seeking to peer with a large one. More to the point, what agreements
occur or fail to occur when one network is in a position to strong-arm
the other and does this diverge from the general case?
That having been said, kudos for the excellent research. As far as
objective numbers go, yours are more thorough than any others I've
seen.
Regards,
Bill Herrin
The data set suffers three flaws:
Depending on your point of view, a lot more than three, undoubtedly.
1. It is not representative of the actual traffic flows on the Internet.
There are an infinite number of things it’s not representative of, but it also doesn’t claim to be representative of them. Traffic flows on the Internet is a different survey of a different thing, but if someone can figure out how to do it well, I would be very supportive of their effort. It's a _much_ more difficult survey to do, since it requires getting people to pony up their unanonymized netflow data, which they’re a lot less likely to do, en masse, than their peering data. We’ve been trying to figure out a way to do it on a large and representative enough scale to matter for twenty years, without too much headway. The larger the Internet gets, the more difficult it is to survey well, so the problem gets harder with time, rather than easier.
That having been said, kudos for the excellent research. As far as
objective numbers go, yours are more thorough than any others I've
seen.
Thank you. We look forward to your participation in the next one! 
-Bill
I think you're over-specifizing Bill's assertion, Woody.
He didn't mean "TCP Flows", I don't think; he was simply -- as I
understood him -- talking about the 40,000ft view of connections between
pieces of the Internet.
I don't expect your dataset to have flow-level data, and I don't think
he did either; it isn't really germane to the conversation we're having.
Cheers,
-- jra
It is not representative of the actual traffic flows on the Internet.
Traffic flows on the Internet is a different survey of a different thing.
He didn't mean "TCP Flows", I don't think; he was simply -- as I
understood him -- talking about the 40,000ft view of connections between
pieces of the Internet. I don't expect your dataset to have flow-level data, and I don't think
he did either.
How else do you get a representative measurement of “actual traffic flows on the Internet?”
We’ve got adjacency information. Telegeography has hand-waving 40,000 ft. flow estimates in the form of different widths of arrows on a map. But if you want to know how large actual flows of data are between two regions of the Internet, and you can’t actually instrument the whole Internet, you need two things: (1) a broad and representative sampling of flow data, and (2) a complete measurement of a few portions of the network that are represented in the sampled set. That gives you a horizontal and a vertical view, from which you can extrapolate to a whole, or any other part, with some minor assurance of reasonability.
If someone has an easier methodology to suggest, that still produces usable results, I’m all ears.
it isn't really germane to the conversation we're having.
I thought I’d made that point?
-Bill
This most likely won’t happen unless it becomes some sort of an international treaty obligation and even then it would end up in courts for a long time. Leaving aside data privacy requirements many carriers have, most companies guard their traffic information rather zealously for some reason.
-dorian