Networking is not especially hard. A great deal of the writing about networking is very hard indeed, and the industry has been letting the two be mistaken for one another for a very long time.
You have almost certainly met the paragraph I mean. It announces a protocol, defines it using three terms it has not defined either, and moves on with the serene confidence of someone who has never once been lost. Beneath it sits a diagram: nine boxes, four clouds, and an arrow whose direction is genuinely anyone's guess. It was drawn by a person who already understood the topic. That is precisely why it explains nothing to a person who does not.
A wall of acronyms is not rigour. It is a description of the subject standing where the explanation should be.
So I am building Daveth for two readers at once, and it is not allowed to trade one away for the other. The first has just opened a terminal and is not yet sure whether the blinking cursor is waiting for them or quietly judging them. The second has kept production networks alive long enough to distrust anything that sounds this friendly, and will go looking for the exact sentence where the explanation gets comfortable and stops being true. Both of them are right to want what they want. A lesson written only for the first is a bedtime story. A lesson written only for the second is a reference manual with nicer margins.
Which is why a lesson here is not permitted to open with a definition. It has to open with the problem that made somebody invent the thing in the first place, because a technology you have no need for is nearly impossible to remember, and a technology whose absence you can feel is nearly impossible to forget.
Then it compares that problem to something ordinary (post, queues, doorways, an argument about whose turn it is to speak) and then, without being asked, it tells you exactly where the comparison falls apart. Every analogy has a cliff edge somewhere. Material that hands you the analogy and then wanders off has taught you a falsehood with a very pleasant surface, and that falsehood will come looking for you during an outage, at an hour you did not choose.
After the model comes evidence. Real output, healthy first, so that the shape of a working system is something you have actually seen rather than something you assume. Then the same output when it has gone wrong, so that failure becomes a thing you recognise instead of a thing that happens to you.
Typing a command proves that you can type. The verification afterwards is the part that proves anything else.
None of this is negotiable, which is an unbearably grand way for me to put it, so here is the plain version. A Daveth page is not finished when the build goes green. A green build means the file has no syntax errors; it holds no opinion whatsoever about whether a human being learned something. A page is finished when a reader who knew nothing can follow it end to end, a reader who knows a great deal cannot put a finger through it, and the learner walks away holding proof rather than a pleasant feeling.
That is the standard I want to be held to. Wherever Daveth falls short of it, that is a defect rather than a matter of taste, and I would like to hear about it.
