Aside from “war is a mere continuation of policy by other means”, one of the concepts most associated with Clausewitz is probably his idea of “friction”. So, if you hadn’t already filled in that space on your Buzzword Bingo card, here you go! That being said, the word “friction” doesn’t actually show up very much here. Which follows, since he’s describing it, not using a word to define itself.
“Everything is simple in war, but the simplest thing is difficult.” This is a really true statement. But Carl expands: “through the influence of innumerable trifling circumstances, which on paper cannot properly be taken into consideration, everything depresses us and we come far short of our mark.”
When running a network, it is the simplest things that can have the weirdest behavior. Domain Name Service (DNS), which is literally what makes the Internet usable by most humans, should be one of the best-behaved protocols at this point. And yet, its utterly incomprehensible behavior at seemingly random moments is a meme.
One of the most frustrating things to hear, whether you are a military leader or network engineer, is “why don’t you just…?” This phrase usually springs from a lack of understanding of how parts interact. And that each of these parts isn’t like a cog in a machine, but an organ of a living organism. You can’t treat each element as an independent variable in an algebra problem. You have to understand “that no part of it consists of one piece.” I highly recommend this post on the difference between “complex” and “complicated” to further explore this concept.
So, we have a system that we depict as a bunch of boxes, individual units, because while all models are wrong, some models are useful. We, as professionals have to understand that even a single piece of hardware is composed of dozens of components, and nudging each of them just a bit can cause some significant changes in behavior. As an example: take a piece of networking gear that has redundant power supplies. It happily works off of one power supply, but redundancy is good, so you plug in the other one to the other PDU (power distribution unit. Basically: a giant power strip. Most racks have two, connected to different circuits) in the rack as is good and proper. Except that circuit is a little bit off, so the second power supply is getting slightly different voltage. Most of your equipment in the rack configured thusly has no problems. Except this one switch. It keeps throwing errors. Have fun hunting that problem down. And I generally find computers and electronics to act more predictably than humans.
This weirdness of behavior, “this enormous friction which is not concentrated, as in mechanics, at a few points, is, therefore, everywhere brought into contact with chance.” In plain English: in a mechanical system you can usually identify a stress point, a component that will wear out and fail before everything else. You can then work to strengthen that component, or make it easily replaceable. Warfare, and computer networks, are not nearly that simple. While you can identify some high-probability events (these are things you work to make redundant, or you avoid those bottlenecks) there are also many low-probability things that will bring everything to a screeching halt. These things aren’t concentrated in any one area, and they can all fail independently. That is, just because five things go wrong today, doesn’t mean you’ve exhausted your bad luck. Each dice roll is completely independent of the previous dice roll, though you seem to have a streak of crit fails. Or critical hits.
So, how do you operate in an environment where things can randomly go catastrophically wrong at any moment? In this environment “with ordinary powers one cannot keep event he line of mediocrity.” The standard answer is “training”. But how do you train knowing that “every war is rich in individual phenomena”? Every incident is unique, and your past experiences may not have prepared you for this exact moment? You do the best that you can with the resources you have.
Carl gives us the rather comedic image of teaching someone to swim while standing on dry land. But even that, going through the motions without water pushing back at you, provides better preparation for being in the water than just walking around on dry land. The best training environment for a new analyst is live data, and real cases. The next-best training environment is generated data based on past events. The next-next-best is puzzles that get the analyst thinking in ways that may be useful when presented with live data. And so forth. While some of these provide better value than others, they all provide more value than not training at all.
When running a network, if you’re doing a good job people tend to forget you exist. Similarly, as a leader, if you’re doing a really good job, everything can be going wrong “while to a distant observer everything seems to be going like clockwork.” How does the leader manage this? “The knowledge of this friction is a chief part of that often boasted experience of war which is required in a good general.” The really good leader “must be aware of it that he may overcome it,” without dwelling on the myriad of ways that failure can happen. (Aside, this answers Angry Staff Officer’s question from the other day about factoring in time for things to go less than perfectly)
“Just as the instinctive judgment which as become almost a habit makes a man of the world speak and act and move only as befits the occasion, so only the officer experienced in war will always, in great and small matters, at every pulsation, so to speak, of the war, decide and determine suitably to the occasion.” There’s no replacement for experience. It takes having done this a few times to feel the ground moving beneath your feet and adjust your balance without really thinking about it.
How can you get this sort of practice at failure without, y’know, actually breaking your network? Black Hills Infosec has come up with a way to inject some randomness into your next tabletop exercise.