An Optimistic View of Net Neutrality

The Net Neutrality argument has been puttering along for a few years in the US, and is now starting to get going over here in the UK as well.

On one side are the ISPs, most of whom are telephone and TV cable companies. They are used to being a choke point between the consumer and the rest of the telecoms world, and their managers all have MBAs. The latter point is important because analysing your position in the value chain is a big part of the MBA curriculum, and everyone knows that being a commodity provider is a bad place to be. So the ISPs are desperate to re-establish their choke-hold, which means doing something that forces the other players to treat with them rather than the competition. Hence they want the right to cut deals with, for instance, streaming video providers. If you want to see good quality TV over the Internet from the Foo Video Company, you have to go with an ISP that has a deal with Foo Video, which restricts your choice as a consumer and therefore allows the ISP to charge more. The ISPs also want to charge Foo Video on the grounds that Foo must want its end users to get good quality video and will therefore pay for the privilege.

Everyone else in the value chain (including the domestic Internet users) is up in arms at this prospect for the excellent reason that if the ISPs can capture some of the money flowing around this part of the economy then there will be less for everyone else. They want the ISPs to be required to stay neutral on the grounds that anything else will increase prices (which, from the point of the ISPs, is the point) and restrict new developments because start-up companies won't have the financial and market muscle to cut the necessary deals.

Unlike many commentators I believe that in the long run this argument will be irrelevant, and that whatever various governments do on the subject will make little or no difference, and I believe this because of Metcalf's Law.

Bob Metcalf's original formulation of his law said that the value of a network is proportional to the square of the number of users. In fact that is only true if everyone on the network is equally interested in communicating with everyone else. In reality that isn't true, but whatever the pattern of communication the value of a network still rises more than linearly with its size, which is what matters.

The History of Metcalf's Law 1: Peering versus Transit

Back in the late 1990s there was a big argument over network peering. There were big ISPs and little ISPs. Big ISPs had all the same costs as little ISPs in running points of presence for modem users to phone into, but they also had to run long-haul bandwidth terminated by big routers. A small ISP could hook into this infrastructure just by connecting to a big ISP, which gave the small ISPs a competitive advantage and left the big ISPs without a business case for running that vital long haul infrastructure.

At the time the only kind of connection between two ISPs was "peering", which means that two ISPs share a connection point where they exchange traffic. So the big ISPs implemented a new policy:

1. Henceforth peering connections would only be made between ISPs the same size, and would only carry traffic bound for customers of the other ISP.
2. Small ISPs could pay for "transit" connections to big ISPs. A transit connection would carry traffic bound for anywhere.

There was a big argument over this change. Small ISPs complained that this would lead to the "balkanisation of the Internet", by which they meant that customers of ISP X might not be able to communicate with customers of ISP Y. But in practice that didn't happen. The peering/transit distinction continues to be the mechanism by which the Internet is organised commercially, but the Internet was never balkanised because Metcalf's Law meant it was never in anyone's interest.

To see why this is, consider what happens to the value of a network cut in half. If the value of a network is purely proportional to the size then you now have two networks worth half as much each; the total value hasn't changed. But Metcalf's Law says that the position is worse: the total value of two networks is less than the value of one big one.

So the only way to make money from splitting a network in two is by capturing value faster than you destroy it. But there just isn't that much loose value floating around the Internet, and all the companies involved found that they made more money by keeping everyone connected. So they did.

The History of Metcalf's Law 2: Walled Gardens

Another bright idea floating around in the early days of the commercial Internet was the Walled Garden. This was pioneered by Compuserve in the days before dial-up Internet access, but MSN and AOL were created with similar concepts. The basic idea was to provide a premium branded information service that would be exclusively available to dial-up subscribers. Internet access was added by these providers later because their customers demanded it, but the business model was based on selling a value-added subscription information service that would be better than the free Internet.

The whole idea flopped. Not enough subscribers were prepared to pay the premium. The basic problem was, and is, that as soon as you put up a paywall between your content and non-subscribers you split the network in two: there is the bit you own, and the rest of the network. Your bit is very small, and Metcalf's Law says that it therefore has a much lower value. So nobody wants it.

Metcalf's Law and Network Neutrality

So what about the network neutrality debate? If you googled for "balkanise Internet" a few years ago you got lots of articles from the 1990s arguing about ISP peering and transit fees. Today the same search yields just as many arguments about network neutrality.

The ISPs want to create a new version of the Walled Garden: information providers will pay to get better links to their end users, and certain protocols will get higher bandwidth than others. And I believe that this will suffer the same fate as all the other attempts to fight Metcalf's Law. An ISP that places artificial boundaries between Internet users (wherever they fit in the Internet's various value chains) is lessening the value of the Internet as a whole to those users. If they make a significant difference to the user's experience then they are offering less value, and will lose market share as a result. If they don't make a difference then nobody will care. Either way its not a paying proposition.

Therefore network neutrality will dominate because anything else is less valuable, and therefore will drive away customers.

There is one application where non network-neutral services may survive for a while, which is TV over the Internet. A DVD-quality video stream is around 5Mbits/sec, or about 2GBytes/hour, which is a bit too high to just transport over the Internet to a home internet connection. Never mind what happens when two people in the same household want to watch different things. So for a while it makes economic sense for ISPs to cut special deals with TV providers. But that's nothing new: the economic model is identical to cable TV. Existing cable TV companies would probably like to keep it that way, but their power to do so is very limited. Sooner or later some combination of improved ADSL, fibre to the curb/home/street and wireless connections will provide commodity bandwidth high enough to carry decent video, and cable TV companies will cease to exist.

Petition on Broadband Advertising

I've become rather depressed about broadband advertising. I found it very difficult to discover the web page describing my own ISPs traffic policy. Furthermore if you compare it to this version from January on the Wayback Machine you can see that the headline speed on the "Large" package has gone up from 4Mbit/sec to 10Mbit/sec, but the evening download limit has merely increased from 800MB to 1,200MB, while extra restrictions have been added starting at 10am as well. Somehow that 10MBit/sec upgrade doesn't feel so generous now: when I want to install Fedora 10 I'd better make sure I schedule the download to start after 9pm.

Virgin Media, in common with most other ISPs, says that the few people who make heavy use of their broadband link reduce capacity for everyone else, and that providing truly "unlimited" service for these few would mean higher bills for everyone. They do indeed have a point. If I want truly unlimited service then I'm sure I can get it, for a price. And the fact is that I probably don't want to pay the price. The occasions when I do want a multi-gigabyte download are sufficiently rare that I can put up with scheduling around Virgin's traffic limits.

So why am I depressed about it? Its because my problem with finding Virgin Media's traffic management policy is not unusual. A sample taken from Google found:

• TalkTalk have a 40GB monthly limit right up there on the front page, although I'm not sure if you can find out how much of your limit you have used this month. What happens when you use it all? Presumably you are cut off until next month.
  


• O2 claims "unlimited" usage, but if you search the site for the word "unlimited" you find that excessive use at peak times will lead to warnings and then account termination. They don't define "excessive" or "peak time". I couldn't find a link to this information from their list of broadband features; it certainly wasn't obvious.
  


• Tiscali have a similar policy, except that after three warnings they limit your peak time speed. They also don't define what "excessive" and "peak time" mean, and they don't say how long this will last or what the speed limit will be. This page was two not-very-prominent links away from the package features list.


• Fast actually sells a range of monthly capacity limits. They warn you by email when you hit 90%, and when you go over 100% they throttle you down to 100kBit/sec. Full marks!

So out of 5 providers (including Virgin) we have only one that makes all the limits and policies clear up-front while three try to hide behind weasel words and hard-to-find web pages. This makes it difficult for consumers to figure out what they are buying. Is Tiscali's unstated policy better or worse value for money that one of the Fast packages? Even if its better at the moment, will it still be so next month? As a consumer I have no way of knowing. Worse yet, if I were a technically naive consumer I might not even realise that the question needed asking. "Unlimited" sounds much better than "40Gb monthly limit".

I think that something needs doing about this. However the Advertising Standards Authority have wimped out: they have declared that "unlimited" actually means "95% of users don't hit the limit in any given month". So its hardly surprising that people are confused.

Therefore I have started an on-line petition at the Number 10 web site. If you are a British citizen or resident then I urge you to sign it. The petition calls on the Government to require ISPs to make all caps and limits on their services a prominent part of their advertising. Only when they do so will consumers have a clear choice between different packages.