Colophon

Source

The worked example numbers the Wikipedia articleOutline of academic disciplines, retrieved 2026-08-17 and licensedCC BY-SA 4.0.

Licensing

Two different things are on this site, and they carry different obligations. The line between them is worth stating precisely, because it is not where most people would guess.

Individual discipline names are not copyrightable — nobody owns “Rhetoric”. What the source protects is the selection and hierarchical arrangement, and an address encodes that arrangement directly. So an address paired with its name is adapted material and carries the ShareAlike obligation, even though neither half is protected alone.

The numbering scheme is a different matter. The division allocation, the alternating branch grammar, the filing rules and the hub-card model derive from nothing — they would work identically on any taxonomy. They are original, and they are not offered under the source’s licence.

ShareAlike does not spread across the site. It applies to adapted material, and an independent work merely displayed alongside licensed material forms a collection rather than an adaptation. On Divisions, the table is BY-SA while the explanation around it is BY-NC-SA — two works shown together, not one work.

Reuse the prose or the scheme with attribution to Adilson Carvalho, zettelkasten.adilsoncarvalho.com, non-commercially. For commercial use, ask. Site code is MIT.

Method

Addresses are generated by a Python script that parses the source outline and emits a single JSON file. The site imports that file and never recomputes an address. The generator asserts that all 2,576 addresses are unique before it writes anything, so a collision fails the build rather than shipping quietly.

Drawer order is produced by a comparator ported to TypeScript and unit-tested against the cases a naive string sort gets wrong. It is also cross-checked against the Python generator’s own ordering across the full data set — two independent implementations, same result.

Known imperfections

One malformed entry

1030/9c4 reads “Mississippian culture* Art History” — an artifact in the source where two entries collapsed onto one line. Reproduced verbatim rather than silently repaired, so the mapping back to the source stays auditable.

37 duplicated titles

The source lists some disciplines under two parents, so the same name appears at two distinct addresses inside one division — Glaciology at 2050/1o and2050/1p1, for instance. Both are valid. Filing practice is to pick one as live and cross-reference the other.

Four addresses need a double letter

Sibling groups wider than 26 overflow past z into aa. Any tool that parses these addresses has to expect multi-character letter segments —3021/50aa is segment 50 then segment aa, not 5-0-a-a.

One structural compromise

Two headings in the source sit a level deeper than the four digits can express. Both are only children, so they take the next sub-discipline number rather than branching — Astronomy is 3012, immediately after its parent Space sciences at3011. It reads as a peer of its own parent, and files exactly where you would reach for it.

Build

Astro, no client framework. The only JavaScript that ships is the filter onthe outline. Type is set in system stacks — a serif for headings, a sans for text, and a monospace with tabular figures for every address, because addresses have to line up in a column to be scannable.

Source: github.com/adilsoncarvalho/zettelkasten.adilsoncarvalho.com