@mkbabb/glass-ui 6.0.0 → 8.0.0

package/src/styles/glass/material.css

@@ -1,500 +0,0 @@
-/* glass/material.css
-   The unified .glass-material atom (AW.W22) + the shared ::before grain/specular wiring across the glass family — ONE source for the three material atoms.
-   Carved from the glass.css cascade root; @import-ed by glass.css in cascade
-   order, each in its own @layer components (isomorphic to the monolith).
-   See docs/precepts/design-idioms.md §cohesion-aware partials. */
-@layer components {
-    /* ── The unified glass material (AW.W22) ──────────────────────────────
-       ONE source for the three material atoms that AV.W15 scattered onto
-       opt-in classes + single tiers:
-         (a) the moving-specular `::before` catch-light (folded here from
-             glass-specular-track.css — that file now keeps only the thin
-             `.glass-specular-track` ALIAS + the brackets that point HERE),
-         (b) the `--glass-edge-light` full-perimeter rim (moved off the
-             floating-only block so EVERY rung carries it),
-         (c) the pointer interaction-light hooks (the `--specular-*`
-             `@property` channel mapped from the consumer's --mouse-x/--mouse-y
-             write).
-       The five ladder rungs + `.glass-card` + the `.dock-icon-button` CONTROL
-       compose `.glass-material`, so a Switch thumb, a glass Button, a Dialog,
-       and a Sheet all read as the SAME material without per-component wiring.
-
-       The `.glass-dock` SHELL is OUT of this group BY DESIGN — it hand-rolls a
-       parallel surface (its own `--glass-*-dock` family + a local rim) the dock arm
-       owns. The COHESION CONSEQUENCE is DELIBERATE: the dock shell carries NO shared
-       edge-gleam / moving-specular `::before` (AZ.W-DOCK-TAXONOMY retired the
-       `instrument-strip` engraved-bezel `::before` — the dock shell now has none).
-       The dock's catch-light lives on
-       its CONTROLS (`.dock-icon-button` et al., listed below + in the opt-in
-       transition block), which ARE in the material group — NOT on the shell plate.
-       Adding a shell gleam would re-introduce glass-on-glass over the controls, so
-       the shell-without-gleam is a NAMED EXEMPTION, gated as `dock-shell-exempt` in
-       proof:glass-cohesion (the only exemption beside the legibility allowlist).
-
-       The selector GROUP is the single source. Adding a surface to the band
-       material is one comma-entry here, not a new opt-in class string. */
-    .glass-material,
-    .glass-wash,
-    .glass-quiet,
-    .glass-resting,
-    .glass-floating,
-    .glass-overlay,
-    .glass-card,
-    .glass-specular-track,
-    .dock-icon-button,
-    .dock-tab-button,
-    .dock-select-trigger,
-    .dock-dropdown-trigger {
-        position: relative;
-    }
-
-    /* ── BG.W-GLASS-CLIP-DISCIPLINE — the ONE paint-box clip register ───────────
-       The scattered per-class clip DIALECTS — `.glass-card` → `contain: layout
-       style paint` and `.glass-btn` → `contain: paint` (both surfaces.css) — are
-       RETIRED into this ONE `contain: paint` register on a NARROWED CONTENT +
-       `.glass-card` + glass-atom selector, so every composing content surface
-       inherits ONE discipline instead of two divergent per-class values.
-
-       THE NARROWING (the blast-radius fence — do NOT re-widen):
-       • INCLUDED — the CONTENT tiers (`.glass-wash`/`.glass-quiet`/`.glass-
-         resting`, the tier the real <Card> composes via
-         surfaceClass('glass','resting')) + the `.glass-card` convenience
-         shorthand + the `.glass-btn` glass-atom. `.glass-card`/`.glass-btn` are
-         the AY.W-A11Y-PERF O-4 nested-backdrop-cost surfaces — a glass Button's
-         own backdrop sample must not escape its card box (proof:nested-backdrop-
-         budget reads `contain:paint` on BOTH).
-       • EXCLUDED — the OVERLAY BAND (`.glass-floating`/`.glass-overlay` — reka
-         Dialog/Sheet/Popover/Command/DropdownMenu/HoverCard) carries NO clip: a
-         PopperArrow / submenu / Select-scroll that does NOT portal-to-body paints
-         PAST the rounded corner and a clip would SEVER it (the Popper-arrow-
-         survives cell).
-       • EXCLUDED — the 4 DOCK CONTROLS (`.dock-icon-button`/`.dock-tab-button`/
-         `.dock-select-trigger`/`.dock-dropdown-trigger`) carry NO clip: the
-         BA.W-DOCK-GEOMETRY freed cross-axis + the 1.1× hover plate + the
-         `--dock-control-safe-inset` must stay un-clipped (proof:dock-plate-
-         clearance).
-
-       `contain: paint` clips DESCENDANTS only — the surface's OWN rim / drop-
-       shadow / focus-ring is NOT clipped (device-verified). It is PAINT-only
-       (never `layout`, which would make the box a containing block for
-       abs/fixed descendants — the pre-empt of the overlay-clip class the blast-
-       radius note names). Each clipped surface carries a `border-radius`
-       (`.glass-card` → `--radius-card`, `.glass-btn` → `--radius-pill`, the
-       ladder rungs inherit the host's rounded box; the `::before`/`::after`
-       pseudos carry `border-radius: inherit`) so the clip FOLLOWS the corner (a
-       clip with no radius trims to a SHARP RECT — the bug proof:glass-clip's
-       radius-alongside arm reds). `overflow: hidden` stays FORBIDDEN (it would
-       clip the focus ring). */
-    .glass-wash,
-    .glass-quiet,
-    .glass-resting,
-    .glass-card,
-    .glass-btn {
-        contain: paint;
-    }
-
-    /* (b) The rim — UNIFORM across all rungs (ONE `--glass-edge-light` token,
-       not a per-rung family). The floating/overlay tiers previously wired it
-       into their box-shadow stack; it now rides the material group as a
-       dedicated `::after` inset ring so it composes alongside (not inside) each
-       tier's own shadow recipe. The `::after` grain overlay already claims the
-       five rungs' `::after`; the rim is folded INTO that same overlay's
-       box-shadow below (the grain `::after` gets the rim inset ring) so no
-       pseudo-element collides.
-
-       ── BE.W5 — `--glass-edge-light` IS A WHOLE BOX-SHADOW LAYER, NEVER A COLOR ──
-       `--glass-edge-light` resolves to a COMPLETE box-shadow layer (the token is
-       `inset 0 0 0 0.75px hsl(0 0% 100% / 0.18)` — geometry + color, glass-fx.css:70;
-       the `.dark` arm re-points it to `--glass-edge-light-dark`). It is consumed as a
-       WHOLE LAYER inside a `box-shadow:` list — `box-shadow: var(--glass-edge-light),
-       <other layers>` (surfaces.css:203/371, stack-rail.css:176). The MISUSE is
-       NESTING it as a COLOR inside another shadow — `inset 0 0 0 1px
-       var(--glass-edge-light)` — which expands to `inset 0 0 0 1px inset 0 0 0 0.75px
-       hsl(…)`: an INVALID shadow value → the WHOLE `box-shadow` declaration DROPS →
-       NO rim, NO drop-shadow (worst in dark, where the rim is the silhouette device).
-       THE RULE: read `--glass-edge-light` as a whole `box-shadow` layer, NEVER nested
-       inside another shadow's geometry. To compose the rim's COLOR (e.g. accent-tint
-       it) use `--glass-rim-ink`/`--glass-material-rim` (rim.css — the COLOR seam),
-       NOT `--glass-edge-light` (the LAYER token). */
-
-    /* (a)+(c) The moving-specular `::before` catch-light — the iOS-26
-       illuminate-from-within behaviour. A masked radial-gradient glow rides the
-       surface at the pointer position (--mouse-x/--mouse-y → the typed
-       --specular-* channel). Folded VERBATIM from glass-specular-track.css so
-       there is exactly ONE specular source. */
-    .glass-material::before,
-    .glass-wash::before,
-    .glass-quiet::before,
-    .glass-resting::before,
-    .glass-floating::before,
-    .glass-overlay::before,
-    .glass-card::before,
-    .glass-specular-track::before,
-    .dock-icon-button::before,
-    .dock-tab-button::before,
-    .dock-select-trigger::before,
-    .dock-dropdown-trigger::before {
-        content: "";
-        position: absolute;
-        inset: 0;
-        border-radius: inherit;
-        pointer-events: none;
-        z-index: 1;
-        /* Map the consumer's pointer write onto the typed properties, ON THE
-           PSEUDO. The `--mouse-x/--mouse-y` write lands on the HOST (inline
-           style) as UNREGISTERED props that inherit down to this `::before`;
-           the registered `--specular-*` props are `inherits: false`, so the
-           mapping is declared HERE where the pseudo owns its typed value. The
-           `var(--mouse-x, 50%)` floor is the centred fallback. */
-        --specular-x: var(--mouse-x, 50%);
-        --specular-y: var(--mouse-y, 50%);
-        /* The catch-light intensity at REST — reads the token cohort
-           (tokens.css §specular-intensity), default 0 so a static, unwired
-           surface is CLEAN and the @property `initial-value: 0` dormancy is no
-           longer defeated by a floor. Hover/active lift it through the same
-           cohort (the interaction-light state machine below).
-
-           BG.W-GLASS-DYNAMICS — the press-COUPLED magnitude (absorbs 13.5, soft-
-           gated by F5.1). The specular reads the ONE `--glass-btn-press-t` spring
-           drive (Button.vue / a dock control writes it off `useSpringPress`), so on
-           RELEASE — after the binary CSS `:active` lets go — the gleam SETTLES with
-           the spring's alive overshoot (the press-swell returns) instead of snapping
-           to rest. Soft-gated: `--glass-btn-press-t` defaults 0 (property-regs.css),
-           so an unwired surface reads `max(rest, 0.16*0) = rest` byte-identical — no
-           hard F5.1 dependency. The `<number> * <number>` calc is engine-safe (NOT a
-           registered var nested in an hsl() alpha — the Chromium-computes-to-0 class
-           the `opacity` path deliberately avoids). The button's OWN active LERP
-           (surfaces.css `.btn-glass:active::before`) still wins while pressed; this
-           carries the general release-settle beneath it. */
-        --specular-intensity: max(
-            var(--glass-specular-intensity-rest, 0),
-            calc(var(--glass-specular-intensity-active, 0.16) * var(--glass-btn-press-t, 0))
-        );
-        /* AX.W52 D19 — a thin, BOUNDED, pointer-anchored EDGE gleam (not the prior
-           full-plate central disc that washed small tiles muddy). The geometry is an
-           EXPLICIT-radius circle bounded by `--glass-specular-size` (36% of the
-           plate) whose falloff reaches `/0` well INSIDE the plate (70%) — a tracked
-           catch-light reading the surface near the lit edge, NEVER a whole-surface
-           wash. The core is genuinely WARM-CREAM (L < 100% so the warm hue survives;
-           the prior `hsl(40 30% 100%)` blew out to white). The `--specular-intensity`
-           multiplier drives the layer `opacity`, NOT a per-stop alpha (a
-           registered-`@property` var() nested in an hsl() alpha computes to 0 in
-           Chromium — the cross-engine-safe path); the per-stop alphas are literals so
-           even the lifted hover/active states read as a lens, not a flash. */
-        /* BB.W-LENSING (scope 3) — the motion-reactive EDGE glint UNIFIED with the
-           centered catch-light disc on ONE `::before`. TWO background layers compose:
-           (TOP) the angle-keyed EDGE-anchored conic sweep — a warm-cream arc that
-           brightens the RIM where the pointer's `--specular-angle` grazes (the
-           glass-THICKNESS read iOS-27 fuses with the catch-light); (BOTTOM) the
-           existing pointer-position centered disc. BOTH ride the SAME `--specular-
-           intensity` cohort (the `opacity` below) + `plus-lighter` blend + warm-cream
-           core `hsl(40 35% 92%)` — NEVER a second color/blend fork. The conic is masked
-           to the rim band only (the `--glass-edge-glint-band` radial mask, scope 3),
-           so the glint is a bounded RIM ring, NOT a whole-plate wash (the AX.W52 D19
-           bounded-gleam discipline). The `--specular-angle` floor `0deg` sits the glint
-           at the top edge when unwired (static rest — no ambient drift, no blowout). */
-        /* BB.W-GLASS-ACCENT — the catch-light CORE COLOR composes the per-instance
-           chromatic accent. The warm-cream `hsl(40 35% 92%)` core resolves through
-           the ONE `--glass-specular-core` var, OKLab-mixed toward `var(--glass-
-           accent)` at `var(--glass-accent-strength)`. This tints the CORE only — it
-           is the COLOR axis; W-LENSING owns the GEOMETRY (the conic angle-keyed
-           edge-glint + the radial disc + the masks + the `plus-lighter` blend, all
-           BELOW, untouched). Tinting the ONE core tints BOTH the centered disc AND
-           the rim glint (they share the core), so the data hue reads at the glint
-           where the pointer grazes — the data-keyed colored catch-light. At
-           `--glass-accent-strength: 0%` + `--glass-accent: transparent` (the
-           `@property` defaults) the mix === `hsl(40 35% 92%)` byte-identical (the W2
-           warm-cream floor). Each gradient stop reproduces its original alpha via
-           `color-mix(in srgb, <core> N%, transparent)` (=== the prior `<core> / 0.NN`),
-           so the geometry + falloff is byte-identical at rest. */
-        /* BG.W-GLASS-DYNAMICS — the iOS-27 backdrop-HUE sample seam (the 2nd of the
-           ≤2 chromatic pairs, beside W-GLASS-ACCENT). BG.W-GLASS-SIGNAL-TRUTH (M9) —
-           ONE hue channel, ONE writer: the seam reads `--glass-ambient-hue` (the REAL
-           observer write — `useGlassBackdropLuminance` writes a sampled `oklch()` modal
-           hue there, tokens/glass.css §BE.W-AMBIENT-TINT), NOT the never-written
-           `--glass-backdrop-hue` the prior comment falsely claimed the observer wrote.
-           So a dock/card catch-light picks up the field hue SUB-PERCEPTUALLY over a live
-           aurora WHERE THE OBSERVER IS WIRED, at the bounded `--glass-ambient-strength`
-           (the companion write-strength knob the observer's target owns). NEUTRAL BY
-           DEFAULT — the `--glass-ambient-hue` registered initial `transparent` +
-           `--glass-ambient-strength` initial `0%` make the OUTER mix
-           `color-mix(in oklab, X, transparent 0%) === X` byte-identical, so an unwired
-           surface reads the warm-cream core exactly (the REFERENCE FENCE: no chromatic
-           pickup until the observer writes the hue). The REST hairline (the `box-shadow`
-           below) reads the RAW neutral `hsl(40 35% 92%)` — NOT this core — so the
-           RESTING hairline is NEVER chromatic (prismatic reserved for WS6). */
-        --glass-specular-core: color-mix(
-            in oklab,
-            color-mix(
-                in oklab,
-                hsl(40 35% 92%),
-                var(--glass-accent) var(--glass-accent-strength)
-            ),
-            var(--glass-ambient-hue) var(--glass-ambient-strength)
-        );
-        background:
-            conic-gradient(
-                from var(--specular-angle, 0deg) at 50% 50%,
-                transparent 0deg,
-                color-mix(in srgb, var(--glass-specular-core) 55%, transparent) 18deg,
-                transparent 60deg,
-                transparent 360deg
-            ),
-            radial-gradient(
-                circle var(--glass-specular-size, 36%) at var(--specular-x, 50%) var(--specular-y, 50%),
-                color-mix(in srgb, var(--glass-specular-core) 50%, transparent) 0%,
-                transparent 70%,
-                transparent 100%
-            );
-        /* BG.W-GLASS-DYNAMICS — the resting NEUTRAL specular HAIRLINE (the read-carrier
-           at the calmer blur). WS3 demoted the blur ladder (quiet/resting → 8px), so a
-           demoted plate needs a thin bright rim to still read as GLASS at rest. This
-           inset warm-cream ring FOLLOWS the rounded box (`border-radius: inherit`) and
-           composites `plus-lighter` with the disc/conic above (a bright additive edge).
-           It is NEUTRAL — the raw `hsl(40 35% 92%)`, NEVER the `--glass-specular-core`
-           (which carries the accent + backdrop-hue) — so the resting hairline is never
-           chromatic (the REFERENCE FENCE; prismatic reserved for WS6). It is a STATIC
-           paint (no transition, no @keyframes → `proof:no-layout-animation` + the
-           0-idle-track discipline hold; the interaction transition stays scoped to the
-           :hover/:active block below). It is the SUPPLEMENTARY rim beside the primary
-           `--glass-edge-light` box-shadow rim — the two carry the demoted-blur read. */
-        box-shadow: inset 0 0 0 var(--glass-specular-hairline-width, 0.75px)
-            color-mix(in srgb, hsl(40 35% 92%) var(--glass-specular-hairline-ink, 70%), transparent);
-        /* The `::before` opacity gates the disc + conic (via `--specular-intensity`) AND
-           this rest hairline. A rest FLOOR keeps the neutral hairline lit at rest (the
-           read-carrier) while the interaction lift still rides `--specular-intensity`
-           (hover 0.10 / active 0.16 both clear the floor). `--glass-specular-rest-
-           hairline` (0.07) is a low whisper — at this floor the interaction disc/conic
-           stay sub-perceptual (well under the AX.W52 muddy ceiling), so it reads as a
-           rim hairline, not a center wash. */
-        opacity: max(
-            var(--specular-intensity, 0),
-            var(--glass-specular-rest-hairline, 0.07)
-        );
-        /* Per-layer masks (the `,`-separated list pairs with the two backgrounds in
-           order): the EDGE conic is masked to the RIM BAND ONLY — transparent across
-           the interior, opaque in the outer `--glass-edge-glint-band` (default 22%) — so
-           the angle-keyed glint reads as a THICKNESS rim, not a center wash; the centered
-           disc keeps its own-radius clip (AX.W52 D19). `mask-image` Baseline 2023-12-07;
-           a non-supporting engine ignores it and the `inset: 0` + in-paint falloff still
-           bound the gleam. */
-        mask-image:
-            radial-gradient(
-                circle closest-side at 50% 50%,
-                transparent calc(100% - var(--glass-edge-glint-band, 22%)),
-                black calc(100% - calc(var(--glass-edge-glint-band, 22%) * 0.5)),
-                black 100%
-            ),
-            radial-gradient(
-                circle var(--glass-specular-size, 36%) at var(--specular-x, 50%) var(--specular-y, 50%),
-                black 0%,
-                black 60%,
-                transparent 100%
-            );
-        /* AX.W52 D19 — `plus-lighter` (additive but HDR-clamped, Safari 16.4+) over
-           the prior `screen` (a non-linear lightener that lifted the whole tile
-           toward white — the muddy-card defect). A non-supporting engine degrades to
-           a plain low-alpha warm overlay (the gleam composes normally — no blend, no
-           blowout), the no-workaround floor; NOT a `screen` fallback. */
-        mix-blend-mode: plus-lighter;
-        /* The moving-specular TRANSITION is NOT declared here. An idle/unwired
-           `::before` is STATIC (no interpolation track), so a bare glass rung / an
-           un-armed Card / a glass Button at rest attaches ZERO animation tracks (the
-           keyframes.js runtime counts 0 idle tracks, not 19). The transition rides
-           ONLY the WIRED + interactive surfaces — the opt-in scoped block below. */
-    }
-
-    /* AY.W-GLASS — the moving-specular transition is OPT-IN (the Card wire-or-omit
-       pattern, generalized). It attaches ONLY to a WIRED surface
-       (`.glass-specular-track` — Card emits it conditionally, the dock controls + the
-       Slider thumb compose it) OR to any glass surface under `:hover`/`:active` (the
-       pointer is present, so the interpolation is legitimately live). An idle/unwired
-       `.glass-floating` / `.glass-wash` / bare-rung `::before` carries NO transition
-       and thus NO keyframes.js track. The §6 register is PRESERVED (still
-       `--ease-standard` on the position-tracked + surface-opacity legs — H-motion §6
-       "position-tracked → --ease-standard"; only the SELECTOR scope changes). The dock
-       controls keep the always-on transition (they are interactive controls; their rest
-       intensity is 0 so they paint nothing at rest, but a control legitimately carries
-       the transition). */
-    .glass-specular-track::before,
-    .glass-material:hover::before,    .glass-material:active::before,
-    .glass-wash:hover::before,        .glass-wash:active::before,
-    .glass-quiet:hover::before,       .glass-quiet:active::before,
-    .glass-resting:hover::before,     .glass-resting:active::before,
-    .glass-floating:hover::before,    .glass-floating:active::before,
-    .glass-overlay:hover::before,     .glass-overlay:active::before,
-    .glass-card:hover::before,        .glass-card:active::before,
-    .dock-icon-button::before,
-    .dock-tab-button::before,
-    .dock-select-trigger::before,
-    .dock-dropdown-trigger::before {
-        /* Smooth the position + intensity between pointer samples (the typed
-           @property regs make the percentage stops interpolable). */
-        transition:
-            --specular-x var(--duration-fast, 150ms) var(--ease-standard, ease-out),
-            --specular-y var(--duration-fast, 150ms) var(--ease-standard, ease-out),
-            opacity var(--duration-normal, 240ms) var(--ease-standard, ease-out);
-    }
-
-    .dark .glass-material::before,
-    .dark .glass-wash::before,
-    .dark .glass-quiet::before,
-    .dark .glass-resting::before,
-    .dark .glass-floating::before,
-    .dark .glass-overlay::before,
-    .dark .glass-card::before,
-    .dark .glass-specular-track::before,
-    .dark .dock-icon-button::before,
-    .dark .dock-tab-button::before,
-    .dark .dock-select-trigger::before,
-    .dark .dock-dropdown-trigger::before {
-        /* Softer over the deep canvas — `plus-lighter` still lifts on dark, so the
-           `.dark` cohort runs a lower hover/active rung (tokens.css §specular-
-           intensity `.dark` arm); the rest rung is already 0 so the dark arm only
-           softens hover/active via the cohort. The rest rung re-reads the (dark)
-           token here so a dark `:root` override resolves. BG.W-GLASS-DYNAMICS — the
-           dark arm mirrors the base press-couple `max()` so the `--glass-btn-press-t`
-           release-settle is NOT dropped in dark (the base opacity floor + the neutral
-           box-shadow hairline are inherited unchanged). */
-        --specular-intensity: max(
-            var(--glass-specular-intensity-rest, 0),
-            calc(var(--glass-specular-intensity-active, 0.16) * var(--glass-btn-press-t, 0))
-        );
-    }
-
-    /* (c) Interaction-light lockstep — hover/active lift the intensity in the
-       SAME rungs the scale-press squish keys off, so material-light fires in
-       lockstep with the press across every glass surface. Settles back on leave
-       via the opacity transition above. */
-    .glass-material:hover::before,
-    .glass-wash:hover::before,
-    .glass-quiet:hover::before,
-    .glass-resting:hover::before,
-    .glass-floating:hover::before,
-    .glass-overlay:hover::before,
-    .glass-card:hover::before,
-    .glass-specular-track:hover::before,
-    .dock-icon-button:hover::before,
-    .dock-tab-button:hover::before,
-    .dock-select-trigger:hover::before,
-    .dock-dropdown-trigger:hover::before {
-        --specular-intensity: var(--glass-specular-intensity-hover, 0.1);
-    }
-    .glass-material:active::before,
-    .glass-wash:active::before,
-    .glass-quiet:active::before,
-    .glass-resting:active::before,
-    .glass-floating:active::before,
-    .glass-overlay:active::before,
-    .glass-card:active::before,
-    .glass-specular-track:active::before,
-    .dock-icon-button:active::before,
-    .dock-tab-button:active::before,
-    .dock-select-trigger:active::before,
-    .dock-dropdown-trigger:active::before {
-        --specular-intensity: var(--glass-specular-intensity-active, 0.16);
-    }
-
-    /* Selected and dragged controls hold the same interaction lens used during
-       press. Semantics choose the state attribute; the material owns its paint. */
-    :is(.glass-material, .glass-wash, .glass-quiet, .glass-resting, .glass-floating,
-        .glass-overlay, .glass-specular-track, .dock-icon-button, .dock-tab-button,
-        .dock-select-trigger, .dock-dropdown-trigger):is(.is-active, .active, [data-active],
-        [data-dragging], [data-state="checked"], [data-state="on"], [aria-pressed="true"],
-        [aria-selected="true"], [aria-expanded="true"], [aria-current="page"])::before {
-        --specular-intensity: var(--glass-specular-intensity-active, 0.16);
-    }
-
-    /* Disabled glass keeps its neutral separating hairline but cannot wake the
-       interaction lens through hover, press, or retained state. Content ink is
-       untouched; each control continues to own its disabled legibility. */
-    :is(.glass-material, .glass-wash, .glass-quiet, .glass-resting, .glass-floating,
-        .glass-overlay, .glass-specular-track, .dock-icon-button, .dock-tab-button,
-        .dock-select-trigger, .dock-dropdown-trigger):is(:disabled, [data-disabled],
-        [aria-disabled="true"])::before {
-        --specular-intensity: var(--glass-specular-intensity-rest, 0);
-        transition: none;
-    }
-
-    /* BB.W-LENSING (scope 2) — the `:active` LENS-SWELL is RETIRED (DDR-LENS-BAKE,
-       glass-refract.css header). The swell drove the displacement `scale` by mutating
-       the `--glass-refract` axis on the `--glass-btn-press-t` press drive (28 rest → 44
-       pressed). That depended on the var()-spliced head/`scale`/tail filter — which never
-       PARSED as a `backdrop-filter` (three tokens: `url("…") <n> url("…")`), so the lens
-       rendered `none` and the swell animated a value that reached NO filter. A CSS url()
-       string token cannot carry a substituted `var()` (the engine does not re-tokenise
-       inside a string), so a dynamic per-instance `scale` is structurally impossible in a
-       single self-contained data-URI — and the single token is mandatory for both parse
-       validity and bundler url()-rewrite safety. The `scale` is now BAKED at the §10
-       default (28); the resting lens read is PRESERVED (and now actually renders), the
-       press-depth animation is the deliberate tradeoff. The press squish + gleam lift on
-       the SAME `--glass-btn-press-t` drive (material gleam, above) carry the press read. */
-
-    /* ── BE.W-CLEAR-VARIANT — the Apple-Clear `surface="clear"` decoration + the
-       MANDATORY legibility scrim ──────────────────────────────────────────────────
-       The 4th surface-axis member (`useSurfaceAxis.ts` maps `clear → glass-clear`).
-       A `surface="clear"` surface paints the maximally-translucent `--glass-bg-clear`
-       (0.58 — the clearest rung the library ships, the album grid bleeds through) AND
-       a MANDATORY legibility scrim IN THE SAME SEAM — a `.glass-clear` cannot exist
-       without it (the Apple Clear contract: near-transparent → the text needs a scrim
-       floor). The scrim DIMS THE BACKDROP (`color-mix(in srgb, var(--background) N%,
-       transparent)` — the A5-1 modal-scrim idiom, painted BEHIND the translucent fill
-       within the box, see the two-layer-background note below), DISTINCT from the W55
-       `--glass-tint-*` PLATE-darken axis (the two legibility axes stay disjoint: the
-       scrim dims the backdrop, the W55 tint darkens the plate). The scrim strength
-       DERIVES from the sampled
-       `--glass-backdrop-luma` (useGlassBackdropLuminance's existing write) so it dims
-       MORE over a bright album grid (luma→1) and relaxes over a dark one (luma→0) —
-       the dynamic legibility that betters Apple's fixed scrim. The `0.5` fallback is
-       the safe mid-scrim when no observer is wired. The plate fill rides the SAME
-       oklab-tint seam so the W55 darken reaches it; the see-through frost is the
-       cross-engine FLOOR (pure opacity + backdrop-filter, no `url()` — Safari paints
-       it via the build-time -webkit- prefix pass). The scrim is a STATIC paint (no
-       layout, no animation — `proof:no-layout-animation` GREEN by construction).
-
-       THE SCRIM IS A SECOND BACKGROUND LAYER, NOT A PSEUDO. The glass family already
-       claims BOTH pseudos on every base rung — `::before` (the moving specular core)
-       AND `::after` (the grain overlay, ladder.css). `surface="clear"` composes ON a
-       base tier (`glass-floating glass-clear`), so its `::before`/`::after` are taken;
-       a scrim pseudo would CLOBBER the specular or grain. The mandatory scrim is
-       therefore the FIRST of two comma-separated `background` LAYERS — painted FRONT
-       (the translucent glass fill) over BACK (the `--background` dim) within the box,
-       so the dim sits between the backdrop-filtered backdrop and the content (a child
-       paints above both). The fill stays translucent so the backdrop still reads
-       THROUGH the dim — see-through AND legible, one element, no pseudo collision. */
-    [data-surface="clear"],
-    .glass-clear {
-        /* BG.W-GLASS-SIGNAL-TRUTH (M6) — the MANDATORY scrim is a STATIC FLOOR the
-           sampled luma can only LIFT, never `var(--glass-backdrop-luma, 0.5) * 40%`
-           whose `, 0.5` fallback is DEAD (a registered `@property` initial 0 —
-           property-regs.css — is never guaranteed-invalid, so the fallback CANNOT
-           fire on a target engine, and the unwired resolve was `calc(0 * 40%) = 0%`:
-           the scrim the canon calls FORBIDDEN-to-omit painted NOTHING). Now: an
-           UNWIRED clear plate paints `--glass-clear-scrim-floor` (12% — the mandatory
-           minimum a scrim-less clear surface can never drop below); a wired observer
-           adds the luma-driven ramp on top (`--glass-clear-scrim-ramp` 28% — floor +
-           ramp caps at the prior 40% ceiling over a bright grid, luma→1). Bare
-           `var(--glass-backdrop-luma)` reads the registered initial 0 unwired (floor
-           alone), lifts CONTINUOUSLY as the observer writes. */
-        --glass-clear-scrim-floor: 12%;
-        --glass-clear-scrim-ramp: 28%;
-        --glass-clear-scrim-strength: calc(
-            var(--glass-clear-scrim-floor) +
-            var(--glass-backdrop-luma) * var(--glass-clear-scrim-ramp)
-        );
-        background:
-            color-mix(
-                in oklab,
-                var(--glass-bg-clear),
-                var(--glass-tint-source) var(--glass-tint-strength)
-            ),
-            color-mix(
-                in srgb,
-                var(--background) var(--glass-clear-scrim-strength),
-                transparent
-            );
-        backdrop-filter: var(--glass-blur-floating);
-        position: relative;
-    }
-
-}

package/src/styles/glass/progress-rail.css

@@ -1,78 +0,0 @@
-/* glass/progress-rail.css
-   The @utility glass-progress-rail — restyles <Progress variant=default> into the deck-position rail. A @utility (not @layer components) so it outranks the variant's baked h-4/rounded-pill utilities.
-   Carved from the glass.css cascade root; @import-ed by glass.css in cascade
-   order, each in its own @layer components (isomorphic to the monolith).
-   See docs/precepts/design-idioms.md §cohesion-aware partials. */
-/* ── Deck-position rail (AW.W16 · AX.W24) ───────────────────────────────────
-   A Tailwind `@utility` that RESTYLES the shipped `<Progress variant="default">`
-   (the reka-ui ProgressRoot track + its indicator child) into the thin
-   bottom-of-deck position rail. It composes the EXISTING fill — it does NOT
-   re-author a second track/range mechanic. Every visual axis is a
-   `--progress-rail-*` token, so a consumer/dark-flip re-tints with zero library
-   edit (`:root { --progress-rail-fill: … }`). The viewport-pinned chrome
-   (`position: fixed`, `env(safe-area-inset-bottom)`, the z-index) is CONSUMER-
-   supplied — this recipe owns only the LOOK.
-
-   AX.W24 — WHY `@utility`, not `@layer components`. The rail RESTYLES the default
-   Progress: it must override the `h-4` height + `rounded-pill` radius utilities the
-   variant bakes. Those utilities live in `@layer utilities`, which ALWAYS outranks
-   `@layer components` regardless of source order, so the prior `@layer components`
-   recipe's `height: 3px` was a silent no-op (the rail painted as a 16px bar). An
-   `@utility` lands the recipe IN `@layer utilities` alongside `h-4`, where a single-
-   class selector wins the geometry by source order — the idiomatic Tailwind-v4 way
-   to author a restyle-class that out-ranks base utilities (the `btn-audacious` /
-   `cartoon-surface` pattern).
-
-   The COLOUR fix is the SEPARATE token-read-at-source path (NOT a layer-juggle):
-   `ProgressDefault` reads `--progress-fill`/`--progress-track` and dropped its
-   `bg-secondary`/`bg-primary` utilities, so this recipe FEEDS those tokens rather
-   than fighting a utility colour. The `--progress-fill` set on the rail root cascades
-   to the indicator child (a CSS custom property inherits), so the indicator reads the
-   override hue. */
-@utility glass-progress-rail {
-    /* The track hairline + zero radius — the rail reads as a thin square sliver, not
-       the variant's 16px `rounded-pill` bar. A single-class @utility ties `h-4`/
-       `rounded-pill` on SPECIFICITY (both 0,1,0) and Tailwind's utility sort can emit
-       `h-4` later, so the bare tie loses. The doubled-class self-reference (`&&` →
-       `.glass-progress-rail.glass-progress-rail`, specificity 0,2,0) wins the geometry
-       DETERMINISTICALLY over the base utilities regardless of emit order — the precise,
-       contained way to out-rank a base utility without `!important`. */
-    && {
-        height: var(--progress-rail-h, 3px);
-        border-radius: 0;
-    }
-    /* Feed the token-read default variant — these route the consumer's
-       `--progress-rail-*` axes into the `--progress-fill`/`--progress-track` tokens
-       the variant reads at source (the cascade-correct colour path). */
-    --progress-track: var(
-        --progress-rail-track,
-        color-mix(in srgb, var(--foreground) 8%, transparent)
-    );
-    --progress-fill: var(--progress-rail-fill, var(--primary));
-
-    /* The fill — the shipped ProgressIndicator child. The indicator paints the
-       inherited `--progress-fill`; the leading-edge glow is re-authored (AX.W24 F2)
-       as a TRAILING right-edge INSET box-shadow that lives INSIDE the indicator's
-       own box. The prior `4px 0 8px` OUTSET rightward shadow extended PAST the fill
-       boundary into the ProgressRoot's load-bearing `overflow-hidden` clip and was
-       eaten — invisible at every value. An `inset` shadow offset leftward
-       (`inset -Npx 0 …`) brightens just inside the indicator's right (leading) edge
-       — the moving position head — and never crosses the fill boundary, so the clip
-       cannot eat it (the clip is never lifted; that would re-break the default
-       Progress look for every other consumer). `box-shadow` is a separate property
-       from `background`, so it composes cleanly over the token-read fill. */
-    && > * {
-        border-radius: 0;
-        box-shadow:
-            inset
-            calc(-1 * var(--progress-rail-glow-x, 4px))
-            0
-            var(--progress-rail-glow-blur, 8px)
-            calc(-1 * var(--progress-rail-glow-spread, 2px))
-            color-mix(
-                in srgb,
-                hsl(0 0% 100%) var(--progress-rail-glow-strength, 55%),
-                transparent
-            );
-    }
-}