An alphabetical index sounds like the simplest structure in information design, and it stays simple for about two hundred entries. Past that, real names arrive — accented characters, mononyms, particles, stage names, non-Latin scripts — and every assumption behind “just sort them A to Z” fails in a different way.
Where sorting breaks first
- Diacritics. Does Ángel file under A or after Z? Naive byte sorting puts every accented character after every unaccented one, which scatters names across the end of the index.
- Particles. Van, de, von, al-, Mc and Mac are handled differently by different national conventions, and sometimes differently for two people with the same surname.
- Mononyms. Performers with one name have no surname to file under and break any schema assuming two fields.
- Non-Latin scripts. Sorting Cyrillic, Devanagari or Han characters into a Latin A–Z requires a transliteration decision, and the transliteration is frequently disputed.
- Name changes. The index must find someone under both names without duplicating the entry.
The standard solution: sort keys
Mature indexes store a separate sort key alongside the display name. The page displays “Ángel Di María”; the index sorts on “Di Maria, Angel”. The two are deliberately decoupled, which solves diacritics, particles and transliteration in one move.
This is why well-built A–Z listings look obvious and are not. A name with a particle — Ana de Armas Feet is the standard case — has to display one way and sort another, and a large index is doing that invisibly for every entry under it.
Redirects do the rest
The second mechanism is aliasing. One canonical page holds the content, and every alternative form — stage name, birth name, common misspelling, married name, transliteration variant — is a redirect pointing at it.
| Variant type | Example pattern |
|---|---|
| Stage name | Performing name → legal name, or the reverse |
| Transliteration | Multiple romanisations → one canonical spelling |
| Misspelling | Common error → correct entry |
| Partial name | Surname alone → full entry, where unambiguous |
| Former name | Previous name → current entry |
The rule that keeps this from becoming a mess: redirects never hold content. The moment two versions of a page both have text, they diverge, and the index now contains two different answers to the same question.
Disambiguation
Every index eventually has two people with the same name. The convention is a qualifier in parentheses — occupation or birth year — and a disambiguation page at the bare name listing both.
The judgement call is whether one is dominant enough to take the plain name. Get it wrong and most visitors land on the wrong person; refuse to decide and every visitor gets an extra click. Large indexes generally use traffic data rather than argument to settle it.
Why flat lists stop working
A single page listing everything is fine at a thousand entries and unusable at fifty thousand — the page is too large to load and too long to scan. The standard response is layering: an A–Z split, then secondary axes such as country, category or attribute, so that the same entry is reachable by several routes.
An entry carrying a diacritic in its display name, such as Natalia Czerwonka Feet, has to be reachable from the plain-ASCII spelling as well as the correct one, which is what the alias layer is for.
The test for a good index
Pick a deliberately awkward name — one with an accent, a particle or a stage name — and try to find it three ways: by browsing the letter, by searching the display name, and by searching a variant spelling. An entry like Anya Sarre Feet, where the person is known professionally by a name that does not follow the usual surname convention, is exactly the case that separates an index that was designed from one that was merely sorted.
If all three routes land on the same page, the index was built by someone who had met real names before.
