Rethinking out loud about infrastructure

One of the most common expressions in geology is “not well understood”. Which is understandable, because most rocks were formed millions to billions of years ago, often under conditions, and in locations, that can only be guessed at. One of the reasons I love geology is that the detective work is of a very high order. The work is both highly scientific and highly creative. Also, it will never be done. Its best mysteries are rooted too deeply in the one thing humans — relative to rock — severely lack: time

Anyway, I’m here to suggest that two overlapping subjects — infrastructure and internet — are not well understood, even though both are made by humans and can be studied within the human timescale. The term “infrastructure” has been in common use only since the 1970s. While widely used, there are relatively few books about the subject itself. I’d say, in fact, that is more a subject in many fields than a field in itself. And I think it needs to be. Same with the Internet. Look it up on Google and see how many different definitions you get. Yet nothing could be more infrastructural without being physical, which the Internet is not.

Anyway, as I write and think about this stuff, I like to keep track of what I’ve already said, even though I’ve moved beyond some of it. So here goes:

More from allied sources:

And now I have to fly to Paris, to have fun at LeWeb. We’ll pick up this and other subjects there.

Tags:

11 comments

  1. Russell Nelson’s avatar

    I’m too lazy to see if I’ve yet pointed you to the idea of “linear infrastructure”. Whenever you look at the history of linear infrastructure (pipes, roads, and wires), you see a common thread: it’s hard to have a competitive market. The first party into the space has the competitive advantage, which they can use to charge monopoly prices. On the other hand, when government tries to run the service itself, it so do inefficiently and poorly. An interesting model is the way heritage railroads in the UK are run: with many volunteers and a few paid staff to do the grunt-work. Some of them run a handful of daily trains, approximating service prior to the Beeching Axe.

  2. Richard Bennett’s avatar

    Time-travel a caveman into modern America and he’ll be amazed at our transportation system. Send him back and he’d tell his buddies something like: “they have things called bus stations, train stations, and airports where you walk inside very clean cave and sit on this thing called a seat. You tie yourself down in this chair thing and it takes you far away. When you leave the cave you’re in a totally different place. Dudes, it’s magic.” Many descriptions of the Internet infrastructure are like that.

    It’s really a communications system, and unlike other communications systems it’s made of diverse elements: fat things like Ethernets, SONETs, ATMs, and odd things like FiOSes, DOCSISes, DSLs, WiMaxes, and Wi-Fis. It doesn’t seem like these things could all be part of a unified system, but they are. And unlike other communications systems that go to great lengths to ensure they provide predictable and invariant service, the Internet has a wierd attitude toward overuse and overload: it says “bring it on, I can handle myself no matter what you do to me.”

    This is what makes the economics of Internetting so brilliant. If we’re willing to give up a little predictability, we can generally have great service for very little money. As more uses migrate off the old-style networks to the Internet, their communication needs come with them, so Internets of the future will be able to meet a wider range of needs without sacrificing economy.

    This infrastructure continuously improves itself, which is very cool, and very different from other infrastructures that really can’t because they’re stuck doing one thing in the same old way by law and regulation.

    That’s the deal with Internet infrastructure. It’s not magic, not free, and not above the laws of physics, but it’s very flexible, very fast, very cheap, and getting more so. It’s not magic, but it’s very, very clever.

  3. Mike Warot’s avatar

    The basic problem with infrastructure is it’s the last thing in the world you want to have to update rapidly, because of it’s massive scope.

    If we could put bundles of single mode fiber everywhere, and only worry about the ends of the fibers, that would greatly help.

    Because we can’t get a do-over ever 2 years, we should design the system with at least 2 pair to every home, business, apartment, etc.

    The base system requirement should be 1 Gbps full duplex per pair.

    We have to assume that the ends can take care of their own filtering, etc… and NOT try to bake it into the infrastructure.

  4. Richard Bennett’s avatar

    It’s silly to build an infrastructure on the assumption that no traffic shaping is needed, and in fact everything we know about network design says that networks in which there is no delay ever need to constrain endpoint access speed to make that happen. This is how you build a telephone network, not how you build a computer network.

  5. Doc Searls’s avatar

    Quick question for the panel from the floor at LeWeb in Paris: would you call the Internet “interstructure”?

  6. Mike Warot’s avatar

    The nice thing about the InterNetwork is that you can always patch around a hole in the existing network. It would be nice if you didn’t have to manually do it, though.

Comments are now closed.