A drawing is an instruction, not a picture

What a drawing promises, and to whom it promises it.

A drawing is not a picture of what a part should look like, it is a set of instructions telling somebody who has never seen the part exactly what to make, down to the specific number that decides whether each dimension has been made correctly or not, and a drawing that only shows what the part looks like without saying so precisely has not done its job at all.

Learning to draw properly, mostly by drawing things nobody could have manufactured from the drawing alone, since every early attempt captured what the part looked like well enough to recognise and left out everything a workshop would actually have needed in order to make it.

Every mark on the page removes a decision

A drawing exists to let somebody who was never present when a part was designed make that part correctly anyway, without ever needing to ask the designer a single question. Every dimension on it states what the part is required to measure, which is a different thing from recording what one example happened to measure, and every other mark on the page (the lines showing the shape, the notes about finish or material) exists to remove ambiguity about what is being asked for. A drawing that shows a beautiful, accurate picture of the part but leaves any dimension unstated or any requirement unclear has failed at the one job a drawing actually has, because a picture only needs to look right, while an instruction has to be followed correctly by someone who has never seen the part at all. The formal system of symbols and rules built up around this, geometric dimensioning and tolerancing, exists purely to make that instruction as unambiguous as possible, closing off exactly the kind of reasonable-sounding misreading a looser drawing would leave open.

A single missing dimension on an otherwise complete, carefully drawn part can stop a workshop from making it at all. A maker asked to guess at even one measurement is no longer following an instruction but making a decision the drawing was supposed to have made for them, and one unmade decision is enough to produce a part built to the wrong size, however carefully everything else on the page was specified.

Why a musical score works and a photograph of a concert does not

Sheet music tells a musician who has never heard a piece exactly which notes to play and for how long, precisely enough that two musicians reading the same score produce recognisably the same performance without ever having heard the original. A photograph of a musician playing shows what one performance looked like, in vivid detail, and yet gives nobody the means to reproduce a single note of it, because a picture records an outcome and says nothing about how to arrive at one. A technical drawing is the score in this pair: it lets a workshop that has never seen the finished part produce it correctly, guided entirely by what the drawing states. A musician handed only a recording could copy the notes reasonably well by ear but would have to guess at anything the recording left unclear, which is exactly the position a workshop is left in by a drawing that only shows what a part looks like.

A score also carries things no photograph could, such as how long a rest lasts or which of two identical-looking notes is to be held, and a drawing does the same with the details a picture hides: which edge a hole is measured from, how smooth a face must be, how far a length may wander before the part is rejected.

Testing a drawing on somebody who was not in the room

Because a drawing's job is to remove decisions, a good one is judged by whether it could be handed to a workshop that has never seen the part and never spoken to the designer, and still result in exactly the part that was intended. Any drawing that would need a phone call to clear something up has, to that exact extent, failed to be a complete set of instructions. This is also why a drawing is never really finished until it has survived contact with somebody who was not there when it was made. A designer reading their own drawing already knows what every dimension is for, and that knowledge quietly fills in gaps a stranger to the part would trip over immediately, so the designer's own careful checking can never fully replace a fresh pair of eyes. The cheapest version of this test is simply to hand the drawing to a colleague, ask them to describe how they would make the part, and listen for the first point where they have to guess.