# Word lists of `wordkey.Generate`

The lists from which `datekeys encrypt -new-words` draws a key of words at
random (spec §38.1: at least 6 words of a public list of 2048 or more). They
are not normative: a reader does not need them, because the key is derived
from the normalized text of the words, whatever list they came from.

| List | Words | SHA-256 | Status |
|---|---|---|---|
| `en.txt` | 7776 | `6d557f0693958fb5e650b68b5bee585eb82cf4da32965505c789e924743bc522` | The large wordlist of the EFF, as published |
| `es.txt` | 7776 | `ff77b487765c000da97cca58fe94a2cdb947303e7a07460614d7d95d800034fe` | Draft, not yet reviewed by a native speaker |

Each list has 7776 = 6^5 words, sorted, so that five dice give a word: each
die minus one is a digit of the position of the word in base 6, the first
die the most significant. The dice 11111 give the first word and 66666 the
last, as in the list of the EFF.

A list changes only with its hash in this file and in `generate_test.go`:
an application that downloads a list pins its SHA-256 and refuses any other.
`wordkey.CheckList` checks a list against the alphabet of its language,
which the code gives and not the list: a word with a letter of another script
that looks the same, such as a Cyrillic `а` (U+0430) for a Latin `a`, would be
typed again with the letter of the keyboard, and the capsule would not open.

| Language | Alphabet |
|---|---|
| `en` | `a` to `z` and the ASCII hyphen of the four compound words of the list of the EFF (`drop-down`, `felt-tip`, `t-shirt` and `yo-yo`), in lower case |
| `es` | `a` to `z`, `á`, `é`, `í`, `ó`, `ú`, `ü` and `ñ`, in lower case and NFC |

## `en.txt`

- **Source:** the large wordlist for passphrases of the Electronic Frontier
  Foundation, by Joseph Bonneau (2016),
  <https://www.eff.org/files/2016/07/18/eff_large_wordlist.txt>, whose
  SHA-256 was
  `addd35536511597a02fa0a9ff1e5284677b8883b83e986e43f15a3db996b903e` when
  it was downloaded on 7 October 2026. The EFF explains it in
  <https://www.eff.org/deeplinks/2016/07/new-wordlists-random-passphrases>.
- **License:** original material of the EFF, which its
  [copyright policy](https://www.eff.org/copyright) licenses under
  [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/), unlike the code
  of this module (Apache 2.0).
- **Change:** the dice number before each word is left out, and nothing
  else: the words and their order are those of the EFF, so the position of a
  word still gives its number.

## `es.txt`

- **Source:** the frequencies of
  [FrequencyWords](https://github.com/hermitdave/FrequencyWords) by Hermit
  Dave, `content/2018/es/es_50k.txt`, counted on the OpenSubtitles 2018
  corpus, licensed under CC BY-SA 4.0. The Spanish Hunspell dictionary of
  LibreOffice (RLA-ES, `es_ES.dic` and `es_ES.aff`) is used only as a filter
  and is not redistributed.
- **License:** this list is an adaptation of FrequencyWords and is licensed
  under [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/),
  unlike the code of this module (Apache 2.0).
- **Method:** the most frequent words that are, in order:
  1. lowercase letters only, of 3 to 9 letters;
  2. a base form of the dictionary: an entry with affix flags, or the
     feminine its flag `G` makes. Entries without flags, which are
     conjugations, plurals and pieces of names, are left out;
  3. not in a short list of offensive or unpleasant words;
  4. not the other half of a pair that differs only in a final `-o` or `-a`
     (`chico` and `chica`): the more frequent stays;
  5. not the same as a word already taken once normalized as in §38.1
     (`papa` and `papá`): the more frequent stays.

  The first 7776 that pass, sorted. `wordkey.CheckList` checks the last rule,
  the characters of every word and the alphabet.