AT&T rigs net neutrality study

We need more bandwidth, right ?

"an internet where all traffic is treated equally would require much more capacity than a tiered infrastructure"

Well duh. Obviously it would need as much capacity as there is traffic. Why is that a bad thing ?

A "tiered infrastructure" under control of AT&T would simply mean that I would get billed for each byte I cause to pass on "their" network, instead of paying a fixed fee per month. Oh of course, that's not how AT&T puts it, but in the end, that's what will happen.

I find it rather galling to see telcos try and find more excuses to bill me for copper that they put in the ground 50 years ago instead of using their immense profits to put fiber everywhere and get a good image for furthering the increasing need in data transfer.

Here's a thought for all those important CEOs : instead of trying to extort money from me by visibly shafting me, be nice and friendly and I'll give you the money you want on my own.

Net neutrality...

can be achived with a tiered network too, but you must make sure that the open internet can be accessed by everyone and paid as a guaranteed bandwith service. Premium content could be introduced but only as a non internet service that has separate bandwith allocated and runs in a separate sub network. This way a provider could charge for a guaranteed 2 Mbit connection where you could access youtube and everything else and offer a package with an extra 4 Mbit connection only to youtube but that connection should be separate so the user could use the 2Mbit link too while they access the 4Mbit extra channel. (so they would have a total of 6 Mbit to youtube if they want to use both channels for watching films but one connection should only use one of the two channels to avoid mixing the two networks)

The idea is that the service providers should switch from overselling their bandwidth to guaranteed bandwidth services, so users could get extactly what they paid for. (this system is already implemented in some parts of europle and japan)

Yawn

It doesn't matter whether they rigged this report or not, they're still right: services must be prioritised and pricing is one way to do this. If the carriers do all the hard work of building the infrastructure then they should be reasonably free to set prices as long as this does not include anti-competitive practices, ie. subsidising the same services themselves. Quality of service is far more important than net neutrality.

Copper

Someone got it right here. Copper. Most of the ISP's in the UK sell '8Mb' services. Read closer and it's 'up to 8Mb', so far so fair. So you install it and check your bandwidth, 5Mb. Or as in my mates house in a village, 1Mb. But you're still paying for 'up to 8Mb'. I don't see any tiering in price structure taking place there. Oh wait, that's because the consumer is paying top price.

They bleat about the cost of the infrastructure and upgrades but I don't see any improvement at my end.

And when a lot of the industry moved up from 2Mb lines a while back, did they automatically upgrade services across the board? Did they f***? In fact I know several people who were still on 2Mb AND paying a premium price until I alerted them, they then hassled the ISP, got faster connections and a reduced price.

Am I missing something ?

Surely the issue is that :

If a carrier believes that they are not making enough money from carrying bits, then they are not charging enough for the connections they sell. Presumably, outfits like Google, eBay, YouTube, etc all pay handsomely for their internet connectivity - so in effect they are paying for the bits they shove into the net. If what they pay isn't enough, then the price is wrong !

Now if it's a case of a third party carrier not making enough from carrying those bits, then they aren't charging enough in their peering arrangements. So if (for example) AT&T don't directly connect Google et al but are complaining about the cost of carrying their traffic, then they should charge more to the carriers responsible for injecting that traffic into their network via peering arrangements.

No I suppose the next argument is that there isn't enough price elasticity in the market to increase prices. Well tough - that's market economics for you. If you can't sell a product at a price that makes a profit then you have to either reduce the costs until it is profitable or simply stop trying to sell that product. Simply whinging about it isn't going to get much sympathy !

Having said all that, I think both sides are wrong on the net neutrality debate - and at least one side (probably both) have simply been distorting the arguments to make it sound worse. I bellieve what most 'pro neutral' folks actually want is that no-ones bits get treated as second class and deliberately shoved onto a slow lane because they won't pay extra - but this is being distorted deliberately as "treat ALL traffic equally" which I think would be bad. On the 'anti neutral' side, they clearly have an agenda to screw more money out of people - as one comment has said, effectively extort money from people "or your site gets crap connectivity".

Geography =! topology

despite efforts of countries like China to firewall off the rest of the world.

Yet now, AT+T seem to be trying to create the First, Second and Third worlds online. First world being corporations, and possibly subsidised americans, second world being feepayers elsewhere, and a thirld world of poor nations and those who can't afford ISP tier rates.

QoS only works until you get to Peering Point

This whole claim that Google should pay ATT for getting Priority for their traffic only is feasible if Google talks directly to ATT. If Google gets their connectivity from Provider A (ie: Not from ATT) then it will flow over Provider A's network until it can get to a peering point where it can be injected in ATT's Network (having possibly first traveled over Provider's B and C's networks (since A and ATT do not directly peer or the routing tables do not select a peering point where they do peer). The way TCP/IP IPV4 (the current version in use) is defined, there is no method to flag the packets as needing special handling or QoS priority (the same thing) end-to-end. The flags that were designated for this purpose are basically only usable for intranet usage [ie: On your own network not for traffic that will go to others networks]. Only by switching to TCP/IP IPv6 is this end-to-end QoS/Prioritization supported. IPv6 is only in its early stages of implementation. IOW: ATT wants to get paid (ie: Charge the provider and the customer) for something they can not deliver since they can only control the packets and treat them differently while they are on the ATT network and they lose control once they get handed off to a peer and they have no control until a peer hands them the packet. Those packet flags I mentioned if used, may/will have different meanings on each network the packet passes though (unlike with IPv6) so as the packet goes from network to network it MAY get different treatment as the flags's meaning gets interpreted differently by each network.

Dear Michael

As an American I would like more information on how to become subsidized as it sounds quite appealing. Mind explaining?

UNev Reno & Rensselaer Poly

Are these insitutions truly hotbeds of expertise in net economics & design? Or did AT&T shop around for individuals with the right prejudices and then choose insitutions that housed them?

Seems to me that if AT&T had really wanted an impartial study, you'd see the report credited to, say, Stanford and Carnegie-Mellon, not a couple of intellectual backwaters.

An analogy would be...

people in West Virginia getting charged less for phone calls 'cause they talk slow.

Valid within its assumptions

I have worked with KK Ramakrishnan and as traffic and queuing experts go, he's one of the best. The study appears to be valid within its assumptions. And its assumption is that you have loss-sensitive flows an non-loss-sensitive flows, which is quite valid. So two ways to handle this are a) handle them differently, or b) overprovision so that nobody gets loss. Well guess what folks -- answer b requires a lot more capacity! I think this is so obvious that it's amazing that people don't accept it -- they're so hung up on the vague notion of "neutrality" that they start denying any math that seems to go against a very fundamentalist "best effort for everything" view.

The real issue around neutrality is not allowing different applications to request different treatment. The problem is when the *network operator* makes that decision. Some networks are starting to deploy Deep Packet Inspection, essentially wiretapping of the customer payload, to do either/both of deciding on how to handle the packets, and how much to charge. For instance DPI vendors have proposed charging a fee for wireless connections to email servers other than the network operator's, because this competes with high-cost SMS. And ATT has talked about taking a value-based cut out of ecommerce transactions passing over their wires! And imagine censoring the news based on the network operators' preference or kickbacks. "No BBC for you, chump, we got Fox!" This is the type of thing that is entirely legal now. Once Big Ed Whitacre at SBC started blabbing about those plans, the spin doctors recovered by trying to distract the conversation onto video prioritization. That's a distraction.

UNev Reno & Rensselaer Poly

I can't speak for UNev Reno, but RPI is a top-notch school. That doesn't mean they can't be bought off, however. Even engineers like money.