properties /measurement

THE PROPORTION THE TWO COMPONENTS ARE COMBINED IN, as the maker writes it - `2:1`, `100:55`, `1:1`, `100:80`. Published as a named field wherever there are two components: Henkel prints `Mix ratio, (by volume) Resin : Hardener` and `Mix ratio, (by weight) Resin : Hardener` on LOCTITE EA E-20HP (2:1 and 100:55) and EA 3450 (1:1 and 1:1); Huntsman publishes a Mix Ratio table for Araldite 2011 (Parts by weight 100/80, Parts by volume 100/100). CARDINALITY MANY WITH A REQUIRED `mix_ratio_basis`, and the requirement is not caution but arithmetic: Araldite 2011's two bases give 100:80 and 100:100 for one product, so a bare ratio is wrong for one of the two readings. A LITERAL AND NOT A TOKEN, on [L11]'s recurrence test read the other way: `2:1` and `1:1` do recur, but the space is unbounded (any maker may publish any proportion) and the values carry no shared meaning to govern, unlike the closed letter-scale of `friction_grade`. Bound to `LiteralCode`, which is `/^\S+$/` - read from `src/formats.ts` at the mint and not quoted from a comment. THE NORMALISATION RULE IS PART OF THIS DEFINITION AND NOT AN IMPLEMENTATION NOTE [A-8]: LOCTITE EA E-20HP PRINTS ITS WEIGHT RATIO AS `100 : 55`, WITH SPACES, AND THAT STRING FAILS THE FORMAT. A seeder copying the sheet character for character will be refused by the gate with no explanation, so the rule is written here - STRIP ALL WHITESPACE AND STORE `100:55`. `LiteralCode` remains the right binding rather than `LiteralText`, and the difference from b154 is real: a ratio genuinely IS one unbroken token once normalised, where eight of nine of b154's specification designations (`API SN PLUS`, `EN 15651-4 EXT-INT CC 25HM`) carry meaning-bearing internal spaces that must not be stripped [H2-f14]. Stated only by a two-part product; a one-part leaves it unanswered, which is the tolerance b154 recorded for `pour_point` on `ServiceFluid`. Concern measurement.