Russ Nelson's blog

[ Home | RSS 2.0 | ATOM 1.0 ]

Mon, 26 Nov 2007

Voting on what's for lunch

You may have heard that "Democracy is two wolves and a sheep voting on what to have for lunch." (It's not from Ben Franklin.) I don't think that's quite accurate. I prefer this one: "Democracy is one wolf and two sheep voting on what to have for lunch -- and lamb still ends up as the entree."

You don't change the nature of things merely by voting.

posted at: 20:59 | path: /politics | permanent link to this entry

Sun, 18 Nov 2007

Fractional Reserve Banking

I'm trying to figure out whether derivatives have the potential to create inflation. In order for them to do that, they would need to be able to create new money where none exists. A lot of people agree with the Wikipedia article on Fractional-reserve Banking. That article says that new money is created when a bank makes out a loan. They're allowed to loan out a large fraction of a new deposit, as long as they keep a fraction on reserve. They make this loan by creating a new account with the loan. That loan can itself be loaned as, as it's considered to be a deposit. And so on and so on until the loan has been multiplied several times.

The problem with that theory is that when somebody takes out a loan, they do so for a reason. Since banks want to be paid more than they pay out, nobody makes money by taking out a loan and leaving it in the bank. No, people spend that money by withdrawing it from the bank. This fellow agrees with me.

So, fractional reserve banking cannot persistently create new money -- no more than can kiting a check create persistent new money. Within a few days, the kite falls to the ground, and what looked like new money was seen to be no money.

So my question is whether a complex derivative can cause money to be in two places at the same time, for a long period of time? I still don't know.

posted at: 07:19 | path: /economics | permanent link to this entry

Why we trade

Russell Roberts attempts to explain why we trade in 1000 words. Let me try to be much more succinct:

We trade because buyers and sellers agree on a price which is respectively lower and higher than the value of the thing being traded.

That may sounds like a Zen koan, in that a price is usually denominated in a value of currency, e.g. dollars and cents. So how can a price be higher and lower? The answer is that the price is the same number. It is the value of the thing being traded which varies. In the eye of the buyer, the value of the thing is greater than the price. In the eye of the seller, the value of the thing is less than the price.

That is why we trade.

posted at: 03:09 | path: /economics | permanent link to this entry

Sat, 17 Nov 2007

Ride starting Sat Nov 17 15:26:37 2007

20.81 km 68265.48 feet 12.93 mi 6306.00 seconds 105.10 minutes 1.75 hours 7.38 mi/hr

Rode up to Norwood today for two purposes. First, to check out an old race track, visible on the DOQ aerial photo. On the ground? Nothing. It is completely grown over, and what isn't grown over is flooded. Thank you, Mr. Beaver. Ran into Mr. Franklin Pierce. No, not the 14th President of the United States, but somebody who lives in the area. Like me, he also has been improving the Rutland Trail.

Second purpose was to see if somebody was logging on the road behind our house. Indeed, yes! And one of the things they've done has been to clear out the brush from the old road going across the wetland. I was able to bicycle down to the level of the wetland, and there's a path wending its way to the creek that runs through the wetland. Unfortunately, the bridge is missing, and it was way too cold to consider going through the water.

posted at: 05:55 | path: /bicycling | permanent link to this entry

Thu, 15 Nov 2007

A market for IPv4 routes

John Curran (which is unfortunately just a stub, and I can't even remember all the stuff John has done, so I can't improve it) read my earlier posting No IPv4 Address Exhaustion, and writes to remind me that an IPv4 address is only as useful as its route, and that a bare IPv4 address is as useful as knowing that I live at number 521.

IPv4 Routing

If you know what an IPv4 route is, skip this section. You know how to get to my house not because you know my house number, 521. You know how to get to it because you know how to get to New York State, and then Potsdam, and then Pleasant Valley Rd. An IP address is split up in a similar manner, using a network number, and then one or more subnets. If you're in Canada, you know New York is to the south or east. If you're in Georgia you may have only a vague idea that New York is Yankee-land up north. That's the start of your route. IP packets travel the same way. Once they get closer to their destination, the subnet is used for more exact routing.

Hosts on the Internet are reachable by other hosts because every host has a route to every other host. The vast majority of hosts are on a network with only one connection to the Internet, so they have a "default route". The router they route to may have its own default route. For example, your PC has a route to your wireless router, which is only connected to your DSL or cablemodem so it also has a default route. Larger routers may have multiple connections. They know which hosts lie in one direction, and which in another and they route packets out the appropriate connection.

The trouble with IP routes is that ultimately, you need to reach a machine which has no default route -- which contains a route to every machine on the Internet. It needs to be able to look up this route in less than a millisecond. Thus, the route table needs to be able to fit into memory. Routers with large amounts of memory and high speed interfaces are not commodity items and are thus expensive.

A market for IPv4 routes

In an economically rational world, everything that has a cost also has a price. We don't live in that world. The people who own the core routers (there are many, owned by different companies) don't get to control the size of the routing table. That gets set by the number of routing announcements. The people who get to make routing announcements run ISPs. So, rather than having a market, everyone involved understands that the routing table needs to be kept small. Nobody wants to be the party that broke "the Internet", so everyone is cooperative. Community norms regulate the size of the routing table.

John's fear is that a market for IPv4 addresses will cause networks to be split up and sold in increasingly smaller chunks. In order to route them, each chunk will need a routing table entry. I accept John's fear as rational. So, I say that any market for IPv4 addresses must include a market for IPv4 routes as well. This makes things more complicated. Let's take an example: Clarkson and SUNY Potsdam. They each have a class B, of which they're using approximately 1/10th of the capacity. They could easily sell 3/5ths of their addresses, but how to get them routed?

Clarkson and SUNY Potsdam have the same network provider (or if I'm wrong about that then in principle, they could, since they're only a mile apart). One way for them to sell their IP addresses is for one of them to renumber to the other's network, share a routing announcement, and then collectively sell the other network along with its routing announcement, the renumberer taking the greater portion for their extra effort.

John's fear is that the community norms will not force this kind of behavior. Once there's a lot of money to be made, some parties may feel free to add "just one more" routing announcement. If the norms are not strong enough to stop this, then nobody will bother to abide by the norms, and the Internet will quickly become unreliable as routes fall off the end of the routing table.

But since routing announcements are 1) public information, and 2) blockable, I think that community norms will suffice to ensure that trade in IP addresses will not result in more routing announcements.

posted at: 20:07 | path: /economics | permanent link to this entry

Wed, 14 Nov 2007

Autoconf brokenness

I believe that autoconf is broken. Autoconf creates a configuration script for a package from a template file that lists the operating system features that the package can use.

The trouble with autoconf is in the configure script it creates. There are several problems with configure. First is that it does standard tests about the operating system, but it doesn't save the result of those tests anywhere. In other words, if you are compiling package A and package B, and they both use autoconf, then both of their completely re-do their configuration tests. If you re-run configure, it remembers the expensive results it got the last time you ran it. But results are not shared across packages.

configure is also used to pass in configuration parameters for the package. These parameters can point to system libraries, or define optional features, or allow for cross-compilation.

configure is also used to detect the presence or absence of necessary system libraries. I've seen it cache that information, and not notice that the library has been installed between runs. I haven't seen that happen lately. Maybe that bug has been fixed?

In all of these cases, the problem is that even the least change invalidates every bit of compilation. Configuration changes aren't tracked between configure and make. When you run configure, it creates a completely new Makefile, so the wise person always runs make clean after running configure.

That's dumb. It's just plain dumb.

The purpose of make is to track dependencies between files. You give it a makefile, and when something is older than something it depends upon, it gets re-built. OR, if something depends on something that will be re-built, it also gets re-built. And so on.

Put configuration changes in individual files, each with its own timestamp. In the makefile, record each configuration's dependency tree. When a configuration entry is changed, that change ripples through everything which uses that configuration, but only those things which use that configuration.

If you object to this, then let me ask you: if you had this, then when would you ever use make clean? make clean is in essence a reboot of the make system. It's an acknowledgement of a bug. So why does every autoconf package support make clean? Simple: because autoconf is buggy.

posted at: 23:46 | path: /opensource | permanent link to this entry

Adify test

Just testing to see how the adify ads will look. This one is wide: This one is tall: This one is more square:

posted at: 22:41 | path: /economics | permanent link to this entry

Sun, 11 Nov 2007

Dying for Pleasure

While I was on a bike ride along the Black River Trail, I heard various loud noises coming from the direction of Fort Drum, just across the river from the old railbed, now a paved trail. Well, it's a paved trail part-way. The pavement and indeed, the trail itself, just stops in the middle of nowhere. The reason for that is that a state employee was killed operating a front-end loader while constructing the trail. They basically walked away from the trail at that point. UPDATE 8/08: The trail will be paved soon

I got to thinking about the contrast. On the one side of the river, we have people training to fight a war. Of course they don't mean to give up their life for their country, but that's a risk. On the other side of the river, you have (in essence) a memorial to a person who did give up his life for his county, building a trail for people's pleasure.

Is the one sacrifice to be compared to the other? We all respect the soldier's sacrifice as "the cost of freedom, buried in the ground." And yet should we respect the worker's sacrifice any less? The concept is less grandiose, dying to build a trail. And yet the trail is used for good benefit on a daily basis by the people of the worker's community.

I'm thinking that taking a risk so that you may please the people around you is a risk worth taking. So it just annoys me when I don't get to see the pictures of Chicago that Doc Searls was going to take. In return for what? An immeasureably lowered risk of dying that morning? Fortunately for you, now, nobody tried to stop me when I took this photo of Central Park.

We should take our pleasure when and where we get it and not worry that life is short and people are fragile.

posted at: 06:53 | path: /life | permanent link to this entry

Sat, 10 Nov 2007

Legislation and Legislation

Perry Metzger points out in an email to me (feel free to email your comments to me; spammers certainly do, and I already hate them; I'm not likely to hate you any more, so don't be shy) that there's legislation and there's legislation. I had written that legislation is an act of man and a law is a fact of nature. Perry's point is that legislation can work against a law, or legislation can work with a law.

Consider that nobody but nobody wants an airplane flying a hundred feet over their house. Not even the most crazed airplane fanatic wants that. Thus your property right includes the ability to control the use of the air above your house. And yet it's ridiculous to claim that that right to airspace extends up to orbital level, giving you the ability to control the flight of satellites you can't even see with the naked eye. Somewhere the law stops.

A society could establish the limit through court cases, allowing people who object to go head to head against people who have a need to use the airspace. That would result in a fairly arbitrary number. But why not pass legislation to decide this number? Whether the legislation chooses 500 or 600 feet is a fairly arbitrary number. However, by passing the law, it allows everyone to enjoy their property without having to establish the number by legal action.

posted at: 21:51 | path: /economics | permanent link to this entry

IPv6 economic nonsense

Geoff Huston writes about ipv6, in the Internet Protocol Journal, "So far the business case for IPv6 has not been compelling, and it appears to be far easier for ISPs and their customers to continue along the path of IPv6 and NATs". This is true, because the magic moment for IPv6 has not yet been achieved. When people want to buy Internet access, they want the whole Internet, not just the part reachable by IPv6. Until an IPv6 address can reach the entire IPv4 internet, it will have zero economic value.

But the real nonsense is in the next article by Iljitsch van Beijnum. He's talking about the current consumption of the IPv4 address space. This is the driving consideration behind IPv6. There's nothing in the IPv6 protocol we need, particularly, besides its 128-bit addresses. They could simply have expanded IPv4 with 64-bit addresses, preserving all the existing algorithms and code with simple changes. Instead of holding an IP address in a 32-bit int, an IPv4-64 could have held it in a 64-bit long. Phil Karn argued for that change, but in the end, lost.

Iljitsch acknowledges that free markets have a role to play in the extension of the IPv4 address space. He objects to this, saying

This scenario has several [two -russ] problems. First, when supply is limited and demand is high, prices rise and hoarding becomes lucrative. So the effect of making address space tradeable could be a reduction of available address space rather than an increase. And certainly, as trading IPv4 space becomes more likely, holders of large address blocks will be less inclined to return them. Finally [Secondly, as there are only two objections -russ], more than half of the IPv4 address space in use is held by organizations in the United States, whereas the developing world has comparatively little address space. The prospect of having to buy address space from American companies that got the space for free is not likely to be popular in the rest of the world.
Emphasis mine.

The first emphasized text, "hoarding becomes lucrative" is complete nonsense. How is something lucrative when you hoard it?? How is that even logically possible? Lucrative derives from the word lucre, defined as monetary gain. To hoard something is to accumulate for future use. You can't both use something and sell it for monetary gain! You either get the use or you get the gain; not both.

He goes on to predict that a market for tradeable IPv4 address space will result in ... no market at all. Well, I'm sorry, but you can't both predict that there will be a market for something, and that nobody will use it. If nobody uses it, then there is no market. That's like Yogi Berra's unintentional quip that "Nobody goes to So-and-so's nightclub anymore because it's so crowded."

The second emphasized text is also complete nonsense. Why object simply because American companies got the space for free? Would Iljitsch also object if American companies had gotten the space for a penny an address? What about a dime an address? What about a dollar an address? Would he also object if the American companies were going to end up losing money? To be fair, he is not obviously objecting himself, but is instead saying that other people will object.

Clearly his objection is not to the price that American companies paid, but to the fact that they can sell them for more than they paid. That's called profit, and it's why companies exist. Yes, most of them aren't in the IPv4 address farming business, but they're also not in the real estate business even though some of them own property much more valuable than the business transacted thereon. Nobody suggests that it's unfair for them to sell their property at the appreciated value.

Sometimes companies acquire something which becomes much more valuable through no action of their own. They are just lucky. In this case, however, these companies acquired IPv4 addresses, and by their use of them, created value in those addresses. So, while Iljitsch thinks he's raising an issue of fairness (rich American companies selling IPv4 addresses at a vast profit to poor third-world ISPs who would otherwise be forced to employ homeless people to beg for IPv4 addresses on streetcorners), he's actually ignoring an issue of fairness. If you own something and make it more valuable, then it's only fair for you to sell it for a higher price.

posted at: 19:24 | path: /economics | permanent link to this entry

No IPv4 Address Exhaustion

In a previous posting about IPv6 economic nonsense, I attacked nonsense as nonsense. But actually, the economics of the article are worse than that. At the end of Iljitsch's article on IPv4 Address Consumption, he suggests creating an anti-market. In order to create a predictable destruction of the IPv4 Internet (which he hopes will be supplanted by the IPv6 Internet), he suggests that address trading should be artificially limited. The purpose of this is to create IPv4 pain so that IPv6 will be seen as less painful.

Economists are shaking their heads at this point. "Use a market!" they cry, "Markets are great at allocating scarce goods!". The problem with Iljitsch's suggestion is that it creates pain everywhere. Not everyone will be ready to switch to an IPv6 Internet at the same time. There will need to be a transition period, where some people will need IPv4 addresses, and some can use IPv6 addresses. How to decide who gets and who goes without? You could do some kind of "desperation" metric, where the people who need them more badly get them, and those who don't, don't.

The problem with trying to centrally control the allocation of IP addresses is that the information necessary to do so is not available centrally. And even if it can be gathered centrally, there is no reason to believe that the metric will allocate the addresses according to the wishes of the authors of the metric. And there is no reason to believe that their wishes correspond to the interests of society at large.

An anti-market throws sand into the gears. If there was, instead, a market for IPv4 addresses, then anyone willing to pay the going price could get an address. At some point, the disadvantages of an IPv6 address in an IPv4 world would be overcome by the price of an IPv4 address. At what point? Nobody can say. The information necessary to predict an end to the IPv4 Internet is inside people's heads. The only way to get it out is to create a market and let individuals decide for themselves.

You see, we will never run out of IPv4 addresses. If there's a market, there will always be somebody willing to sell for some price. We will know that we have achieved an IPv6 Internet when the price of an IPv4 address drops back to zero. The way to get there from here is not to artificially create a scarcity of IPv4 addresses. Because, after all, that is the problem, right? How do you make a problem better by making it worse? Shades of medieval doctors' bleeding of patients! Shades of the Vietnam era "We had to destroy the village to save it"! We laugh at these ideas now -- so save time and laugh at Iljitsch's idea -- quickly, before somebody takes him seriously and implements it!

UPDATE: IPv4 routing needs to be part of what gets traded in the market.

posted at: 17:48 | path: /economics | permanent link to this entry

Fri, 09 Nov 2007

300,000 Meters per Second

Anil Dash says "Laws are ours, people -- they're not carved on stone tablets.".

Sort-of. What he says is 90 degrees out of phase with the problem. There is law, and there is legislation. They aren't the same thing. The law is the real thing. You know, like "300 million meters per second -- not just a good idea, it's the law". Laws are discovered, not made. A law might be "people won't produce some kinds of content unless you pretend that it remains their property even after they sell a copy." Legislation is just an attempt to write down a law.

So, we can decide what the legislation should be. When the legislation doesn't follow the law, we can change it, but changing laws (300,000 m/s) is a little harder.

posted at: 17:24 | path: /economics | permanent link to this entry

War on Drugs

A local person, Heather Thomas, wrote a plaintive letter to the editor a week or two ago. She complains that the person who robbed her with a weapon was only getting probation. Why is this person getting probation? Well, the county jail is full, and if we want to put anybody else in jail, we have to pay another county to jail them. So, somebody has to get probation, and we can't be letting the non-violent drug offenders get probation. So criminals walk the streets while druggies whose only crime is trying to change their mental state (or help somebody else do that) are getting a free vacation in jail.

Attempting to put all the drug dealers in jail is simply not possible. There is a demand for their job function, so the only effect of jailing somebody who has taken on that job is to create a job opening at a higher pay rate.

The War on Drugs is a War on Economics. You can ignore economics if you want. You can even fight economics. But economics is going to win every time.

posted at: 06:43 | path: /economics | permanent link to this entry

Mon, 05 Nov 2007

Status report

Status report on the free chordites over in my '770' category. Readers of the main blog may be confused, but pyblosxom allows me to create multiple rss feeds. The 770 category gets syndicated on Planet Maemo. But if you only ever look in the Chordite category, you won't see it.

posted at: 21:38 | path: /chordite | permanent link to this entry

Free Chordite status

I thought I'd report on the status of my offer of two free keyboards. Curiously, I only received three keyboard requests. I meant to impose a barrier to entry, but maybe not that high a barrier!

I decided not to choose, but instead to make three keyboards. I have ordered and received the electronic parts. I've decided that while I might be able to produce keyboards from the hand photocopies the winners sent in, it's too uncertain a process. Still tinkering with the design, but I think I've settled more on a "kit" for a keyboard.

The trouble with the Chordite is that it really needs to fit your hand. And yet, as a portable device, it needs to be sturdy. Adjustable yet fixed. Malleable yet unchanging. This is not a new problem for people to have faced. Screws, nuts and bolts, glue, clay, plaster, wood, plastic, metal, and rock are all substances which can be changed and yet which are sturdy.

Hands are variable in two ways that matter: in finger length and spacing and in palm width. The chordite needs to be stretchy in both those dimensions, and yet, if you drop it, it shouldn't fall into pieces. It needs to be lightweight so that you can carry it. In order to make all this work, I think that I'll put the switches on little PC boards interconnected with 18 gauge copper wires, covered with a layer of polycapralone. The copper for stiffness, and the polycapralone for sturdiness.

I've done some testing, and an ordinary hair dryer puts out enough heat to soften a fairly thick layer of polycapralone. Once softened, the keys can be moved around, with the copper wire keeping the keys in position while the polycapralone is soft.

Once I've made the PC board with the bluetooth module at the heart of this keyboard, I'll ship these to the lucky winners. We'll see what they have to say. If it's not good, then back to the drawing board.

posted at: 21:34 | path: /770 | permanent link to this entry

Archives

My contribution to the Chuck Norris never/always meme:

Chuck Norris is never wrong, even when he changes his mind.

posted at: 20:47 | path: /life | permanent link to this entry

Hospitals forced to close?

So, people think that our health care system would be improved if we only got rid of all that nasty profit by having the government run it all. Maybe, but the way it would be "improved" is by having less health care available. Case in point (for anybody who actually thought that they were buying their health care in a free market): New York State has a Commission on Health Care Facilities in the 21st Century (which is such an unwieldy name that everybody calls it the Berger Commission). You know, "Commission" like in "commision of a crime". But I editorialize too early.

The purpose of this commission is to reduce the amount of health care available to YOU, the buying consumer, by closing hospitals.

Feh on faux free market health care. This is a socialized system ... which of course I do not defend because it is obviously socialized badly.

And if anybody still thinks that health care operates in a free market, try going to a doctor and buying health care. You know, just like you go to McDonald's. You get your treatment (hamburger and fries) and you pay your bill. Only, you can't just pay your bill, you also have to pay a New York State surcharge. Why are you paying this surcharge? Because ... you are ACTUALLY PAYING FOR YOUR HEALTHCARE. You must be some sort of rich person! If you were truly deserving, you would be on medicare like any sane poor person is, so NYS charges you extra for paying in cash.

posted at: 06:38 | path: /economics | permanent link to this entry

Made with Pyblosxom