A weave is not drawn. It is executed. On the simplest loom the warp threads are divided into two groups; one group is lifted, the shuttle passes through the opening, the groups swap, and the shuttle comes back. Lifted or not lifted, once per pass — a binary decision repeated a few thousand times, and the fabric is the output. Once you see a cloth that way, a patterned silk stops looking like a picture and starts looking like a run of instructions.
The three structures and what they cost
Nearly all woven cloth is built from three interlacing patterns. In plain weave each weft passes over one warp and under the next. In twill it passes over two or three, with the point of crossover shifting each row to make a diagonal. In satin the weft floats over four or more, with the crossings scattered so no diagonal appears. The variable is float length: how many warp threads a single weft crosses before it is locked down.
Everything about the cloth's character follows from that number. A plain weave is locked at every crossing, so it is firm, matte, hard-wearing and creases readily. A long float leaves more of the yarn lying flat and unbroken on the surface, so it reflects light as a continuous sheen and drapes softly — and it snags, because those long exposed lengths are exactly what a fingernail catches. Satin's lustre is bought with durability. This is not a matter of fibre quality or finishing; two cloths from the same silk, woven differently, will disagree about almost everything you can feel.
The loom hardware sets the limits. The number of shafts, or heddles, that can be lifted independently determines how long a pattern repeat can be. Two shafts give plain weave. Four give twills and small repeats. Beyond that the weaver runs out of pedals and has to find another way to address individual threads — which is where the loom stops being a frame and becomes a machine with a memory.
The pattern as stored instructions
For figured silk, each warp thread across the width has to be told separately whether to rise on a given pass. The Chinese drawloom solved this by separating the two jobs. A drawboy sat above the loom and pulled groups of cords in a fixed order; the weaver below threw the shuttle and beat the weft. The cords were not attached to the design as decoration. They were the design: a bound sequence, called the pattern draft or flower book, that could be lifted off one loom, stored, copied, sold, and mounted on another. Change the sequence and the same loom weaves a different silk. That is control information held outside the machine, which is the property that makes a program a program rather than a setting.
The Jacquard mechanism of the early nineteenth century, with its chain of punched cards, is the end of this line rather than the beginning. It mechanised the drawboy's job and made the stored sequence durable and cheap to reproduce. It is often described as an ancestor of computing, and the claim is reasonable if it is kept narrow: a stored, repeatable sequence of binary decisions driving an output device. It is not reasonable if it is stretched into a claim that a loom is a computer. A loom has no branching and no arithmetic. What it does have, unambiguously, is a readable memory.
Why the warp is treated differently
Warp threads are under tension from the moment the loom is dressed. They are lifted and rubbed against the heddles and reed on every single pass, thousands of times, while the weft is laid down once and left alone. So warp yarn is sized — coated with a starch or similar film that lays the loose fibres flat and sacrifices itself to abrasion — and it is wound under even, carefully controlled tension. Weft yarn is usually not sized at all, and its tension matters much less. A beginner who uses the same yarn for both and sizes neither will break warp threads all afternoon without understanding why the weft is fine.
The reed, a comb of fixed spacing, decides how many warp ends sit in a given width. Its dent count is a hardware constant: you cannot decide half way through a piece to make the cloth denser. Width, thread count and the basic grain of the finished textile are settled before the first shuttle is thrown, by a piece of metal with slots in it.
What happens after the cloth comes off
Cloth shrinks when the tension comes off. A weave that looked even on the loom tightens as the warp relaxes and the weft crimp increases, so the design has to be calculated for the finished state, not the state you can see while working. This is why woven patterns are specified with a repeat and a density rather than a drawing: a circle drawn on paper will come off the loom as an ellipse if the warp and weft densities differ. Traditional figured silks are full of motifs that read as circular precisely because someone compensated for that ratio in the draft.
Selvedge is the other place the plan shows. It is not a hem added afterwards; it is where the weft turns around and goes back, and it has to be built with a slightly different structure to stop the edge pulling in. A fabric whose selvedge puckers is telling you the weft tension was wrong for the whole piece, even if the middle of the cloth looks perfect.
Reading a woven pattern
Hold a brocade up to a raking light and look at the long floats. You can see where a colour was carried across the back instead of being interlaced, which is how a weaver keeps a motif crisp without thickening the cloth. Look at the repeat: its width is a multiple of the number of independently controllable ends, so the size of a motif is evidence of the machine that made it. Look at the selvedge to see how the turn was handled. None of this is visible in a photograph of the pattern, and all of it is the part that took the skill.
A printed fabric puts a picture on a surface. A woven fabric builds the picture out of the same decisions that build the cloth. There is no layer of image on top; the structure and the design are one object. That is why a figured silk can be read from the back, and why the oldest surviving patterns can be reconstructed from a fragment: the information is in the interlacing, not in a pigment that might have faded away.
The long consequence
The useful idea here is not that weaving anticipated computers. It is that a craft can reach a point where the difficult, valuable thing is no longer the material or the hand skill but the stored sequence of decisions — a thing you can carry, copy, and run again somewhere else. Once that exists, the workshop has two products: the cloth, and the pattern. It took the rest of industrial history to work out how much the second one was worth.