I'm not clear on the claim here. It seems to be, "You guys are using facts to support a position the other guys disagree with, so don't be surprised when they start directly attacking facts and the gathering of facts." I agree that this is typically what happens. I'm not so sure that it's fair to say that both sides are doing equally bad things when it happens, though.

If Netflix split its traffic among a bunch of Tier 1 providers, Verizon would still end up in the same situation. The same number of bytes would be going through the core of its network, and it would have to buy the same amount of hardware at the edge of its network to handle the incoming data. Verizon would enjoy no hardware or overhead savings from that arrangement just as it incurs no extra cost from it all coming from one place. The only difference is which ports get plugged into which cards. The bottom line is that Verizon downloads more than it uploads to all carriers everywhere and no amount of accounting will change that.

If Netflix split across multiple carriers, it would make it harder for Verizon to demand payment, though. It's easy to isolate one juicy target's data and sell them "access" to your customers if that customer's data all comes through one port set of ports that you can just leave unmaintained. If Netflix balanced its load across a bunch of providers, the only way Verizon could do that is to explicitly throttle or drop Netflix packets. But then it would be obvious what was really happening. Verizon doesn't have a problem with Netflix. Verizon doesn't have a problem with unbalanced traffic at all. Unbalanced traffic is part of Verizon's nature. Verizon is just trying to push its own costs off to other providers, unless we're to believe that the whole Internet should pay Verizon for the privilege of connecting to Verizon.

I think you are demonstrating that Netflix breaks the typical peering business arrangements between transit carriers. This is exactly the issue that Verizon and other big ISPs are trying to force.

I don't think it's Netflix that is breaking the business model, though. The Internet used to be more symmetrical when it was academic institutions and government, but that's just not what the Internet looks like anymore. Now it's largely companies with servers sending data to consumers who want data. The old model requiring symmetry just isn't going to hold, Netflix or not. Netflix is an easy target because they're the biggest, but without Netflix, other providers would take their place and provide that content. The "source" would diffuse across more Tier 1 providers, but the destination bottleneck would remain the same. Maybe that would be better because it would require that ISPs realize that the problem has nothing to do with Netflix and everything to do with how humans are now using the Internet (and how ISPs encourage them to use it!).

Further, the symmetrical model likely never held for consumer ISPs. Since AOL was the Internet for most people, end consumers have generally downloaded more data than they've uploaded. Consumer ISPs who are shocked to find that their traffic only goes mostly in one direction aren't going to find a remedy that balances traffic, and pretending that data costs more to push bits uphill than downhill isn't going to change that. It makes no more sense than charging more for 1s than 0s and having the guy who sent extra 1s pay the difference every quarter. The only reasonable long term solution is for each provider to charge their own customers for the cost of the infrastructure they use regardless of which direction the bits flow.

They will be forced to either extract money either directly or indirectly from Netflix or from their end-users. They obviously prefer the former and I agree with them as Netflix is the one most profiting from the massive load that they are causing.

I think this is where we disagree. I think that Netflix and Netlfix viewers are both profiting roughly equally from the arrangement and they each pay their own provider for access to a mutually beneficial link. Netflix isn't imposing data on anybody any more than Netflix customers are an imposition on L3. The difference between L3 and Verizon is willing to charge its customers to cover its costs and Verizon would prefer to have somebody else's customers cover theirs. That's an understandable desire, but not a reasonable one.

Basically, Netflix pays L3 a HUGE amount of money for their access and L3 wants their peer networks to upgrade and continuously carry this huge, unbalanced load without the downstream networks seeing one red cent of that money.

Let's imagine that video streaming was totally symmetrical and Verizon customers sent a byte to Netflix for every byte they downloaded. Would that reduce the cost to Verizon and suddenly make no cost peering equitable again?

Let's say Netflix continues growing by leaps and bounds and absolutely dominates as the source of traffic on the Internet, even more so than it already does. L3 gets paid more and more by Netflix for their access bandwidth while Verizon gets absolutely nothing extra but is required to carry more and more load from L3. That can't work or will require last mile providers to continually raise their prices to upgrade their networks due to Netflix.

That sounds like a sensible model to me. L3 incurs extra costs from Netflix and passes those costs on to Netflix. Verizon incurs extra costs from its customers and passes those costs on to its customers. There seems to be good symmetry there. The source of this problem seems to stem from the idea that all networks need to be balanced all the time. Sometimes traffic flows just one way and that's OK as long as the people who want that traffic transferred pay for it.

I wouldn't have had a problem with Verizon working out a deal with Netflix if it could have been done minus the extortion tactics. Cutting out the middle man can almost always benefit both parties. Unfortunately, when you allow extortion to succeed, you end up with more extortion down the road. I'm especially sensitive to it with monopoly providers like consumer broadband ISPs in the US. If you let a natural monopoly forget that it should be curteous and fair because it should be grateful for its sweet monopoly gig, things can get out of hand fast. If broadband was a truly competitive market, I'd say, "Hey, let them fight it out. Verizon will figure out who they work for." But Verizon has clearly gotten to the point where customers are taken for granted and they can be treated like product instead.

Why? It seems like both parties on both ends of the wire want those bits to be sent. I don't see how the direction of data flow is a useful metric at all. If we suddenly reversed everything and watching videos required that customers upload massive amounts of data to Netflix instead of downloading it, would the cost to Verizon suddenly drop and the cost to L3 suddenly skyrocket? Does the direction of data flow thing hold all the way to the endpoint? Should Verizon be paying its end users for the "right" to dump data they requested into their houses?

That one really blew my mind. I don't see how it's any different from, "We're going to let you rot in prison until you take us to the body, even though we don't have the evidence to convict you without it," which is pretty much exactly what the Fifth Amendment was supposed to prevent.

If 94% of academic economists have fudged things to make their papers look better...

Actually, 94% engaged in some form of "unaccepted" behavior, a category which appears to include such sins as self-plagiarism.

That's why I'm saying that SF Park should simply go back to using its sensors and letting the price float. That will put this app out of business and make the whole point moot. Everything starts to function optimally and there's no need for us to set up police stings to ensure that nobody is selling spaces. Setting the price equal to a sensible market price does away with the secondary market entirely.

I can see other cities getting upset over this, but SF has the tools to stop it cold. They have the most advanced parking metering system in the country. If they don't use it properly, that's a policy problem.

Actually, I think I can phrase this more clearly without bothering with any of the nitty gritty details:

Street parking is a shared public resource. If a shared resource is cheap or free, people will hog that resource. That's fine until we run out of that resource. Then, resource hogs crowd out others and you get a lot of people who don't get to use that resource at all. That's where putting a price on scarce resources comes in. It makes people think about their usage and make rational choices about whether it's important for them to keep using it while others wait their turn.

As you increase the price, fewer people will hog the resource. More and more people will get a chance to use it, and they will use it only for the time they really need it. Yes, you can set the price too high and end up with the resource being underutilized, but that's not the problem San Francisco has. With properly computer-controlled pricing, they can avoid ever having that problem. With proper pricing, you get the largest number of people getting a chance to use the shared resource, and they use it when it's important to them, not just because it's cheap and convenient.

If that's what I meant, that is what I would have said. I did not. You keep making things up and then expecting me to defend them.

There's a reason I used a question mark in my attempted restatement of your position. I'm trying to clarify Because the phrase "ten percent of your users" is kind of ambiguous from the perspective of how a time shared resource is utilized.

Let me rephrase more precisely: "Number of occupied parking spaces" and "time spent in a space by the median parking consumer" are both random variables whose statistical properties change during the day. What does it mean to lose ten percent of your users in that sense? Does it mean that there will be ten percent fewer events of the type "person enters and leaves a parking space?" If so, I disagree with the notion that properly set prices will cause this to happen. If something else, please state it clearly.

More to the point: What are the properties of a set of parking spaces when they're used at maximum efficiency? Is turnover maximized? Is the time the average space spends in the "empty" state minimized? Is time spent looking for a space minimized? I would argue that high turnover and low search times while keeping most of the spaces full most of the time should be the goal. If search times are high, prices are too low. If search times are low because the spaces are mostly empty, prices are too high.

They're not going to feed the meter for another hour unless they really need it, and at that point effect 1 comes into play.

So that's effect number 1, which you and I agree on. You just ignored the more basic effect of (2): that increased prices will reduce the number of hours people buy in the first place. Do we disagree that (2) will happen at all, or do you think that the effect of (1) will dominate? Because in the actual real measured world, (2) dominates (1).

If I have paid an extreme amount (in my opinion) for a certain time on a meter, then I am less willing to just walk away from that investment.

You're also less willing to feed unnecessary amounts of money into the meter in the first place.

If I pay a quarter for an hour at a meter and my business is done in ten minutes, then I don't feel bad about just leaving, opening up the space for the next user. If I pay a dollar for the same amount of time, I'm more likely to see 50 minutes left on the meter when I get done and think "I'm already here, I might as well do something else."

You have a very strange notion of how parking meters work. I've never seen a meter that charges $1 per hour and has a minimum 1 hour purchase. At every meter I've ever seen, you can purchase a few minutes. If you know that you're likely to spend 10 minutes in the store, why would you buy an hour of time? Maybe you would if it only cost a quarter. But if it cost, say $10, you'd probably do what most people do and pay for $2.50 for 15 minutes, do your business, and get out. That's how parking meters really work, especially in a city like SF with the most advanced meters in the country, and especially when you're implementing congestion pricing with the explicit goal of decreasing loiter times.

It is simply absurd to price something to deliberately reduce demand and then deny that you've reduced demand, or at least tried to.

For somebody who screams about "making things up" you've certainly jumped to a weird conclusion about what I'm saying. Raising the price will reduce demand (supply, actually, according to the classic microeconomics terminology) in the sense that it will reduce willingness to occupy the space for a given amount of time. That will have a few effects:

1) Increased turnover.
2) Because of (1), we get increased probability that at any point in time, there will be an empty space.
3) Because of (2), we get decreased time spent looking for spaces, which reduces traffic congestion, driver frustration, and uncertainty about whether you'll be able to find parking when you need it.

If the price is set properly, the spaces should still be mostly full most of the time, so we're not wasting parking spaces by leaving them empty. We've just decreased the amount of time any one person spends in a space, which seems to be what you and I both want. If your thought experiment has increased prices decreasing turnover, it's conflicting with real world experiments, so there's probably something wrong with it.

Your assumption seems to be that a public resource must be priced at a rate to limit demand to what is available, thus optimizing return on investment, not just to cover the costs of providing that service.

It's not at all about optimizing return on investment. It's about efficient allocation of scarce resources and the problems that are created when people try to overconsume those scarce resources. Overcrowded parking is a huge expense in dense cities because people create congestion while driving around looking for spots. The SF Park system was able to dynamically set prices so that there was at least one open space for every N spaces. It kept turnover high and it ensured that people who really need spaces will be able to get them without much cruising. It also distributed parking more uniformly--dynamic pricing encouraged people to park in less busy areas instead of constantly driving around the most congested ones.

With dynamic pricing, really busy spaces cost a lot, keeping people there for short periods of time. Any jackass who tries to occupy the "last" space in order to ransom it off will find that he's paying an exorbitant rate for it and that he's competing with empty spaces that pop up all around him because of the increased turnover. It becomes a money-losing activity.

The number of users of parking spaces will be many times the number of spaces. Each space will be used multiple times a day.

You can add "at any given time" to that to make the statement an accurate model. And assuming you're able to change pricing dynamically (which SF is), it's the correct model to be using. If you can't change pricing dynamically, you're in a bit of a pickle because the price at the beginning of the workday will obviously be totally wrong at, say 10:30pm.

If you force ten percent of those users to go elsewhere because the price is too high, you've lost ten percent of your users.

When you say "lost ten percent of your users", do you mean that if we randomly sample the number of spaces occupied, we'll find 10% of them empty? What's the "correct" number of occupied spaces, given the fact that having drivers driving around without access to empty spaces comes with a cost?

If you make the price high enough on an hour meter, you will increase the number of people who will not simply drive away from time left on the meter. They've paid for an hour, they might as well use it. That reduces the number of users as well.

This, I think is where your model breaks. You've just described a market in which raising the price increases the amount of the resource a consumer is willing to consume. Those markets exist, but they're extremely rare exceptions, and I don't think street parking is one of them. The real model goes more like this:

1) As you say, people who have time left on the meter might be marginally more likely to use it instead of leave early. That has a minor negative effect on turnover.
2) People will put much less time on the meter to begin with because spending time parked in an expensive space costs money. They'll plan to get in and out, minimizing their exposure to meter costs. This has a significant positive effect on turnover.

The net effect should be that (2) dominates (1). Increasing the price should increase turnover significantly. That's what reserach shows, and it's what basic economics predicts.

And, of course, you will lose completely those who would have tried coming downtown to shop if the cost of parking had not become too high to justify it.

By going the other direction, you lose customers who would be willing to pay to park but who don't venture into the area because, "parking is a nightmare." There's no free lunch there. Creating a situation in which a resource is used to its maximum and is rationed by a mixture of luck and waiting in line deters people who aren't willing to drive around in gridlocked streets for an hour looking for a space just as much as a few extra bucks at the meter deters people who don't want to pay a few bucks.

Can you show me where I said it did? Your straw man is very flimsy.

I was mistaken. Most people erroneously think that the number of parking space user-hours isn't limited by the number of parking spaces and that they can somehow get more user-hours out of a space by making the space free. Your error was in thinking that lower prices increase parking space turnover. That's also wrong.

You're right, that isn't the problem here. The problem here is inflating the prices and driving people away, not trying to attract more. And the original problem is increasing the price by running a private auction for a public resource.

The very existence of that auction is due to the fact that SF is underpricing its parking spaces. With their advanced meters, they could very easily just let the price float to market rates. That would have a few effects:

1) The excess revenue earned by the auctioneers would go to the city.
2) Parking would be maintained at optimal density and turnover.
3) This app would go away completely and we could all sleep soundly knowing that nobody had an incentive to ransom spaces, and SF wouldn't have to spend a penny on policing the issue.

Right. That's just dumb. If parking is nearly full but with spaces open here and there, pricing is already optimized. People who are just casually looking for parking are clearly deterred enough to avoid it when possible, most of the spaces aren't being wasted, and there's still enough parking for people who really need it to jump in without cruising for 30 minutes first. That's the holy grail of parking rates. Jacking up the rates won't get you more revenue and lowering rates won't appreciably increase the number of people parking. Anybody who cranks the knob over in either direction clearly doesn't understand how these things are supposed to work.

I guess they could always make more money by lowering rates to fill all of the spaces and then setting really short limits on how long you can park for in order to increase parking ticket revenue (I've seen this in a lot of cities). One parking citation pays more than a full parking space all day, right? But that only works in cities where dicking around with parking is a luxury they actually have. In major urban areas like SF and NY, parking is a serious business. If you're optimizing for anything other than the maximization of space utilization and the minimization of search time, you're causing all sorts of other ancillary problems that just aren't worth whatever revenue you think you're getting.

The very act of pricing public parking to reduce demand means you are reducing demand BY CHOICE, so you really can't argue that you haven't reduced the number of users.

That's not really true. The number of users equals the lesser of the number of people willing to pay for the spaces and the number of spaces. When it's busy (like we're talking about with congestion pricing), the number of spaces is the limiting factor. Increasing the price doesn't reduce the number of cars in spaces until you crank the price way up. It just reduces the number of people cruising around and queueing up for spaces.

Conversely, "free parking" doesn't mean you can cram more cars into the same number of spaces. If you've got empty spaces, sure, you can get more people in by lowering the price. But that's not the problem here.

SF is apparently no longer using the sensors, which is the root of the problem. When the sensors were there, the system was able to keep prices high enough to ensure that there was at least one empty space on most blocks. Under that regime, holding the last space "hostage" would have been prohibitively expensive. They'd essentially be paying the ransom rate in order to demand the ransom.