Russell's Blog

New. Improved. Stays crunchy in milk.

Net Stupidity

Posted by Russell on April 25, 2006 at 7:36 a.m.
"It's not about the customer, you see, it's about their profits," Kossak mcjoan argues. That couldn't be more correct, though probably not in the way it was meant.

You see, I am an AT&T shareholder. I'm also a Comcast shareholder. I am very interested in the profits of these companies. I want them to make great steaming heaps of money because this will become my money. That's why I think it's absolutely essential to preserve network neutrality.

Why, you might ask? Won't AT&T and Comcast make more money if they can charge Google to access their customers? Sadly, No! (apologies to Seabas, Brad and Gavin)

The value of a network is not simply the cost of of the equipment in it. To understand what a network is worth, and thus what you can charge people for the privilege of using it, we're going to have to dip into utility theory.

Before we get into that, fathom a few simple facts:

  • A single telephone is utterly worthless.
  • A heap of telephones and wires in a sack is worthless.
  • A telephone network that connects people who hate each other is worthless.
Utility theory isn't state of the art economics, but because networks are complex objects, utility theory is really the only place to start. It's easy to see how, for example, a cup of coffee or a banana or a back-massage is valuable. When you have one, you're happier. Utility theory allows you to very formally rank your preferences among choices by assigning each choice a certain number of "utils" (yes, they really call them that). You can then attempt to find the best choice by maximizing the number of utils with the right choice.

It's easy to see that a hypothetical person might be happier with a banana, a cup of coffee and a back massage than they would be with two of the three, or with none. Now, say this person is not a morning person, but is nevertheless awake in the morning. A cup of coffee and a back massage might be a lot more valuable to them than a banana and a back massage. Now say you have a bunch of each item (just pretend a back massage is an "item"), and a bunch of people. Some people might have different preferences, and thus different utility functions -- say your group includes a diabetic, a Mormon, and guy with a really bad sunburn. There should be some way of distributing the items such that the total utility is maximized.

So, how happy will you be with one network verses another?

Say we have a very simple telephone network. You pick up the phone, you dial a number, and someone else on the network picks up. You talk about something interesting, and then hang up. What's that worth?

If you're the only person on the network (there's only one phone), the utility of the network is zero. You can talk to yourself without a telephone for free. If you need to talk to yourself, then perhaps the novelty of adding a telephone to mix might be of some interest to you, but let's just pretend you don't talk to yourself on the phone.

If there is one other person on the network, then the network earns one util by allowing you to call that person. As you add people to the network, the value increases with the number of people you can call. However, you aren't the only person on the network, of course, so the network has the same utility for every other person on the network too.

[image from WikiPedia]

The utility function of the network scales with the square of the number of users, but to make sure our function is zero when we've only got one user:

U(n) = ( n * ( n - 1 ) ) / 2

This is called Metcalfe's Law.

Now the Internet isn't like the telephone network. You can do a lot more than simply call one other person at a time. You can read news from news sites, talk with twenty-three of your friends on Instant Messenger, make a VoIP call, post some trollish comments on some blogs, check your personal email, your work email, your email from your college account, and the email account you gave to that cute so-and-so at the bar the other day, upload photos to your photo gallery, download viruses and spy-ware that will slowly cripple your computer (automatically!), leech music from 300 strangers, download the latest episode of West Wing from your mom's TiVo, stream NPR, and fill your iPod with a podcast of someone screaming incoherently in German. And you can do it all at once.

Clearly, such a network has a lot more utility than our simple telephone network. The Internet is about groups of people communicating, not pairs of people. So, we need a new utility function that captures this. The utility function for a network like the Internet is:

U(n) = 2^n - n - 1

This is called Reed's Law, after David P. Reed. Reed helped design TCP/IP.

Indeed, Reed's law may even be an underestimate of the utility because it doesn't take into account the fact that the Internet is simply a bit-moving mechanism, and that the real services people use, which usually follow Reed's Law, run on top of it. It's really an ur-network, and we can build whatever we want on top of it.

However, there is an underlying assumption in both of these models: That any user can connect to any other user, without preference or penalty. Reed's Law and Metcalfe's Law are ideal cases in which the network is utterly agnostic about who connects to who. The ideal cases have the maximum utility, and the less ideal they are, the less utility.

In the case of a telephone network governed by Metcalfe's Law, the non-ideal behavior might be that calls to more distant users cost extra. This may seem somehow more "fair," but the simple fact is that it reduces the value of the network over all. There is a very, very good reason cell phone companies often offer free long distance.

The scaling of Reed's Law suggests that networks that reflect this model would be extremely sensitive to violating the connection-agnostic assumption. Indeed, that stands to reason. If you suddenly find out that you have to pay extra to use Google, you're not allowed to make VoIP calls, advertisement agents will join your private IM conversations to push their wares, that your friends on other networks have to pay per message to send email to you, and that a random smattering of sites are mysteriously blocked, then your network connection isn't going to be worth very much to you, is it?

Indeed not. How many such restrictions and penalties would it take for you to regard your connection as significantly less valuable? Probably not very many. And if people regard the connections as less valuable, what will happen to the price they're willing to pay?

The existence of the Internet has added trillions of dollars of value to the world economy. AT&T and Comcast are asking Congress for permission to slash the value of their own capital investments. They aren't being greedy. For God's sake, please, yes! Let them be greedy! Let them squeeze every last goddamn penny from their network assets!

This is the point where the the Left gets things wrong. Yes, sometimes greed is a vice, but sometimes it is a virtue. It is profoundly dumb to cast the net-neutrality debate in terms of greedy network operators versus virtuous Google, eBay and their own (always virtuous) customers. If it were really true, then of course they should try to collect as much money as possible. But that's not what's going on here, and it's not why people are upset. People aren't upset because they're going to get charged more; they're upset because this will wreck the value of the single most valuable technological and cultural asset on the planet.

AT&T, Comcast, Verizon and Time Warner are not being greedy. They are being idiots.

Non-citizen Patriots

Posted by Russell on April 10, 2006 at 11:02 p.m.
I knew that a lot of soldiers in the US military are immigrants, but I here are a few facts that I discovered today that are rather interesting:
  • All men of military age are required to register for Selective Service, including illegal immigrants.
  • One in ten soldiers who have died in Iraq were immigrants.
  • About five percent (!!!!) of soldiers in the US armed forces are illegal immigrants.
  • The first solder to die in Iraq was Marine Lance Cpl. Jose Gutierrez, who was an illegal immigrant from Guatemala.
People enlist for a lot of reasons. Whatever those reasons, if you're willing to risk your life by serving in the military, I think it's safe to say that you're pretty much comitted to the idea of America. Whatever their politics, no one is going to let people shoot at them for the sake of something unless they believe in it very, very strongly. Some conclusions we can draw:
  • Many immigrant soldiers who died in Iraq enlisted as soon as they arrived. These people fought and died for a new home they were never allowed to enjoy, and they died for freedoms and principles from which they never personally benefitted.
  • Illegal immigrants make up less than three percent of the US population, and yet comprise five percent of US armed forces. That means they are about twice as likely to enlist as a citizen.
Could it be that illegal immigrants are more serious about their patriotism that citizens? If the attendees of the march on Washington today are any indication...

...maybe that's not such a crazy question after all.


Posted by Russell on April 10, 2006 at 11:01 p.m.
In creating a new database for Rails hacking, I accidentally reset the password for my Typo database. After much scrutiny, it was impossible to tell that the problem was caused by Typo no longer being able to log into its database. There was no connection error in any of the logs (that I could find after days of looking).

I only noticed after I tryed to log into the database manally to inspect the tables for damage. So, not only did Typo not generate an error, but it kept serving cached pages without comment.

Dear Typo: Work or don't work. Half working is worse than broken.