Hacker News Reader

Comment by mikeash | Hacker News Reader

By "ongoing costs" (maybe the wrong term) I mean costs related to the amount of traffic.

The number of trucks or airplanes you need to run on any given day depends on how much traffic you carry. The amount of fuel you need to buy then depends on that. So does much of the maintenance you do, and how much you pay the workers. A postal network carrying zero traffic will be a lot cheaper to run than one carrying lots of traffic. The analogy to the internet is not very good. A network link with a 100Gbit capacity costs about the same to run for a day whether it carries a petabyte or nothing. A postal route with a million parcel/day capacity costs way more to run for a day if it actually carries a million parcels than if it carries nothing.

View on HN

> The number of trucks or airplanes you need to run on any given day depends on how much traffic you carry.

That might be how a delivery service works, but it's not how the post office works in almost all the countries in the world. Including, I suspect, the USA.

Post office networks are actually run as networks. They have drivers who collect from post boxes at regular times, sorting offices that work round the clock, postmen who do regular delivery rounds, and so on. The cost of running the network is, over the short term, almost impervious to the amount of traffic.

(Over the long term, of course, you restructure the network to cater for things like population changes -- again, just like the Internet.)

What you should realize is that a lot of people used to think like you before 1840, which was why the Penny Post was such a big deal. Network thinking made mail dramatically cheaper and led to a massive expansion in communications, and the whole system went global in a very short time.

None of that would have happened under the old system.

Isn't that just because volumes are highly predictable in the short term? It's not that the marginal cost per item is near zero, it's simply that the number of items is just about the same each day. It's still meaningful to talk about the cost of delivering a single item, and the strangeness of having a price that's well below that cost.

Yes, volumes are predictable, and no, it's still not really meaningful to talk about the cost of delivering a single item. You have to switch to network thinking -- and I'm sure there's plenty of C19th analysis to support it.

The real cost is the cost of running the network, not the cost of delivering an individual item. Think like that and you'll fail to understand why it's the same price to deliver a letter from Manhattan to LA as it is from Manhattan to Manhattan.

In this case, the problem is not the cost of an individual item from China. The problem is that the pricing incentivizes the sending of large volumes of items from China (where postal prices are low) to the USA (where delivery costs are high).

Unfortunately those prices were negotiated before (I assume) anybody realized it was going to be a problem.

On the good side, the USPS charges high prices to send items to countries with low delivery costs, so it's swings and roundabouts. If the traffic flows between China and the USA were equal, it probably wouldn't be a major problem.

> Think like that and you'll fail to understand why it's the same price to deliver a letter from Manhattan to LA as it is from Manhattan to Manhattan.

Um, they're the same price because it's mandated to be such by law. Go ask a company whose prices aren't set by the government, like UPS or FedEx, and they'll quote you very different prices for those different services.

As for the rest, if the cost of additional items is near zero then why is it a prove that the pricing incentivizes large volumes sent to the USA?

Volume is predictable but not the same.

Mail volume varies easily by a factor 3-4 between days.

The cause would be high volume things would be things like utility bills that usually go to everyone inside an area on the same day or within a few days.

>Over the long term, of course, you restructure the network

And this is definitely happening with the USPS: a few years ago some people I know leased a local post office building that had been decommissioned due to decreased mail volume.

Once you need to send a truck to someone's house, have someone get out and drop off one letter then adding a second letter has a tiny extra marginal cost. Further, there are some fuel savings from sending an air craft 1/4 full vs. 3/4 full but again there minimal in comparison to keeping the aircraft flying in the first place.

Thus, from the post offices perspective they are much better off with a steady stream of traffic than random spikes. Which is a large part of why the post office is willing to deliver spam aka junk mail so cheaply.

Sure, but once the truck is full, sending one more letter needs an entire second truck. Average this across the system, and a route's costs per parcel are roughly equal to (cost of running one truck) / (number of parcels the truck carries).

Nagle's algorithm[0]. So long as you haven't recieved a "Where the hell is my mail" phone call, just keep adding stuff to the second truck until that second truck fills up, and send it later.

[0]: https://en.wikipedia.org/wiki/Nagle%27s_algorithm

The normal solution to that is just wait a day to deliver some mail. With some intelligent routing you basely never send a truck out that's not close to normal capacity.

I'm sure that a router sitting idle is going to consume less electricity and require less cooling than one operating at full load.

Reply on news.ycombinator.com