Fable
The copyist cut the long scroll into pages that could be found
The archive held a batch of very long files. Some went from city gate repairs to bridge taxes, others from caravan disputes to dock regulations. The curator knew these files were important, but every time someone asked a small question, he had to spread out the entire scroll.
The earliest approach was to store whole volumes. This was the least trouble, and the files were complete. But when a reader asked "can a temporary shelter be put up on the south side of the East Bridge," the retriever could only haul over an entire volume of the "Annual Chronicle of Streets and Bridges," ninety percent of which was irrelevant.
Someone found it too long, so they cut each scroll into fixed lengths. Every ten feet, one section, all neat and tidy. For the first few days, speed increased. Then the problems came: a key rule often had its first half in the previous section and its second half in the next; readers saw only half the meaning.
So the copyist switched to cutting by heading. Each time he came to a subheading, he started a new page. This was better than cutting every ten feet, but some files were written messily, with few headings; some headings looked fine, yet underneath were mixed three completely different problems.
A young scribe began by reading the content first. He discovered that the place that truly needed to be cut was not where the ruler reached, nor necessarily where the heading was, but where the meaning changed boundaries: from bridge structure to taxation, from the original text of regulations to adjudication cases, from background to exception conditions.
He also left a small overlap with adjacent pages. Because some issues straddle the boundary, cutting them off completely would leave the next reader without context.
Later, the director ruled: before a long scroll enters the archive, it must first be cut into pages suitable for questioning and retrieval. Pages too large bring back too much noise; pages too small break the meaning apart; cut the boundaries wrong and the answer is skewed from the very start.
Readers still saw the curator's answer, but what really determined answer quality was often the page-splitting work done before those questions arose. The archive staff then understood: material cannot be used just by putting it in. Long knowledge must first be cut into units that can be retrieved, understood and cited. Later, when the archive reviewed several wrong answers, it found that many errors were not the curator misreading, but pages cut awkwardly. Some people got only the rule, not the exception; others got the case but missed the preceding scope of application. The curator then required page-splitters to write down the reasoning for boundaries, and regularly re-cut pages that were retrieved but useless, and pages that should have been retrieved but weren't. Page-splitting went from odd job to a craft that determines retrieval quality.