ADR-0006 field-night palette strategy
#Context
field-night exists to preserve dark adaptation. Rod photoreceptors, which carry scotopic vision, peak in the blue-green middle of the spectrum and fall to near insensitivity at long wavelengths — the basis of the red-light convention in astronomy, marine, and aviation night operations. Every emitted pixel in this theme is a wager against the user's night vision, and the palette strategy decides the currency the wager is paid in. Two candidates were named at Phase 0: strict long-wavelength-only, or a narrow-gamut multi-hue palette. The ramp generator cannot be designed until one wins: the first is a one-dimensional lightness axis at a locked hue, the second is a constrained two-dimensional gamut, and they demand different generator shapes.
#Decision
field-night is strict long-wavelength-only. Every emitting token in the theme resolves to a single locked hue angle in the red-orange band of OKLCH, with chroma defined as a generated curve and lightness as the only semantic axis. The exact hue angle is a generator parameter chosen with the ramps in this phase, orange-shifted rather than deep red so the palette stays legible to protanopic vision. Hue carries zero information in this theme — which is not a concession but the enforcement mechanism for the field-first rule: since every attention-carrying component must already express its meaning through shape, position, weight, and pattern, a single-hue theme is standing proof that it does. field-night is where the field-first claim is tested on every render, not asserted in a guideline.
#Rejected options
Narrow-gamut multi-hue (a small set of desaturated hues, for example red plus amber). Retains a trace of hue reinforcement and reads as less austere. Rejected on three grounds: shorter wavelengths stimulate rods and erode the adaptation the theme exists to protect, so every added hue is paid for in the theme's own purpose; the reinforcement value is near zero by construction, because the field-first rule already forbids meaning from depending on hue; and "narrow" requires a cutoff — which hues are inside the gamut — that would be an unprincipled constant with no defensible derivation.
Achromatic (grey-only at minimum emission). The simplest generator and zero chroma to reason about. Rejected because dim white light still carries the short-wavelength energy that costs adaptation fastest; at equal legibility, long-wavelength light spends less of the user's night vision than grey light of the same apparent brightness. Achromatic is the right answer for a degraded rendering fallback, not for the theme.
User-selectable hue within a permitted band. The strongest accessibility argument, since protanopic users have reduced sensitivity to deep red. Rejected as a palette strategy because a theme whose palette varies per user cannot ship build-time receipts: contrast and emission checks would have to run per configuration, and an unverifiable configuration is exactly what the enforcement stack exists to exclude. The protanopia concern is answered inside the decision instead — the locked hue is orange-shifted, and lightness, never hue, carries the signal.
#Consequences
The ramp generator gains a per-theme hue-lock mode: field-night ramps are one-dimensional lightness curves at a fixed hue with a generated chroma cap, while the interior themes use the full generator. APCA runs against the locked palette like any other theme. The theme-completeness check is untouched — every semantic token still resolves, it resolves to a hue-locked value. The field-night emission ceiling, already the lowest in the system, is derived jointly with these ramps this phase. Touches the Contrast and Theme completeness rows of the enforcement table; adds no new check. ADR-007 is independent of this decision.